FEDERAL COURT OF AUSTRALIA
Trivago N.V. v Australian Competition and Consumer Commission [2020] FCAFC 185
ORDERS
Appellant | ||
AND: | AUSTRALIAN COMPETITION AND CONSUMER COMMISSION Respondent | |
DATE OF ORDER: |
THE COURT ORDERS THAT:
2. The appellant pay the costs of the respondent, to be assessed if not agreed.
Note: Entry of orders is dealt with in Rule 39.32 of the Federal Court Rules 2011.
ORDERS
VID 150 of 2020 | ||
| ||
BETWEEN: | TRIVAGO N.V. Appellant | |
AND: | AUSTRALIAN COMPETITION AND CONSUMER COMMISSION Respondent | |
JUDGES: | MIDDLETON, MCKERRACHER AND JACKSON JJ |
DATE OF ORDER: | 4 november 2020 |
THE COURT ORDERS THAT:
1. To the extent that it is necessary, leave is granted for the appellant to appeal from the orders of the primary judge made on 28 February 2020 at Melbourne in proceeding VID 1034 of 2018.
2. The appeal be dismissed.
3. The appellant to pay the costs of the respondent, to be assessed if not agreed.
Note: Entry of orders is dealt with in Rule 39.32 of the Federal Court Rules 2011.
Table of Contents
THE COURT:
1 The appellant, Trivago N.V., operates an online search and price comparison site for hotel accommodation. The Trivago website swiftly aggregates hotel room offers from many online sites and displays those options to the user.
2 The issue in this appeal is whether the information Trivago conveyed or, importantly, did not adequately convey to consumers who used the Trivago website between 1 December 2016 and 13 September 2019 (the Relevant Period), either alone or together with Trivago’s advertising, was misleading or deceptive or likely to be so.
3 At trial the primary judge held that it was. For the reasons that follow, his Honour’s conclusion was substantially correct.
4 The appeal (and therefore these reasons) are factually intense. It follows that it is also necessary on the appeal to address the evidence in detail. But for two evidentiary sub-issues, only one substantive point of law arises.
2. THE NATURE OF THE TRIVAGO WEBSITE
5 Central to this appeal is an understanding of the Trivago website as an aggregator of hotel offers, as distinct from Online Booking Sites whose websites allow consumers to directly book hotel rooms. Online Booking Sites reached arrangements with Trivago in order that their hotel offer prices would be displayed on Trivago’s website so as to reach potential consumers who might then click through from Trivago’s website to the Online Booking Site and make a booking.
6 At dates relevant to the litigation, the Landing Page of the Trivago website enabled consumers to input a city (or region) in a search bar, to select their desired dates and room type, and to choose the filter they wished, if any, to produce results. When a consumer initiated a search of this nature, the Trivago website would display an initial set of search results in response to the selection criteria inserted by the consumer (the Initial Search Results Page). This page displayed a list of hotels. Each hotel listing was displayed in substantially the same format. To the left of the screen in each instance was an image relating to the hotel being advertised and certain information about it. On the right were multiple offers for each hotel from identified Online Booking Sites. One such offer was displayed on the far right side in green text in a standalone column (for example, “Booking.com AU$299”). This was known as the Top Position Offer. On occasions a box was displayed underneath the Top Position Offer, showing that the Top Position Offer had particular attributes, such as free cancellation or free WiFi, pay at the hotel (rather than in advance) etc.
7 Additionally, there were three offers along with each of the Online Booking Sites’ names, displayed in smaller sized grey text directly to the left of the Top Position Offer (for example, Expedia AU$299, Wotif AU$299, Destina AU$420). These were described at trial as the Second, Third and Fourth Position Offers. Underneath the Second, Third and Fourth Position Offers a “More Deals button” was displayed which, when clicked, would show, by way of a “slide-out” function (More Deals slide-out), additional offers for the selected hotel. Initially the More Deals button indicated the number of further offers within the More Deals slide-out. After about 5 December 2017, the More Deals button was changed to read “More Deals from AU$...”. It would display the price of the cheapest deal contained in the More Deals slide-out, for example, “More Deals from AU$227”. The cheapest deal contained in the More Deals slide-out was sometimes at a lower price than the Top Position Offer.
8 There was a “sort by” filter option which appeared above the listings on the Initial Search Results Page. The default filter was “our recommendations”. Other filters included “price only”, “rating only” and “distance only”. Additional filtering tools were available to consumers who wanted to exclude offers on the Initial Search Results Page including a price per night scrolling bar, where consumers could set the maximum cost per night. They could also filter by accommodation type, for example, hotel or house/apartment; and by hotel location, for example, number of kilometres from the city centre.
9 Prior to about 20 November 2018, above the Top Position Offer, for some hotel listings on the Initial Search Results Page, there was a higher priced offer from another named Online Booking Site in red strike-through text. This was known as the Strike-Through Price. From on or about 20 November 2018, the Strike-Through Price was replaced by red text without a strike-through, in a smaller text size referred to in the litigation as the Red Price. The Red Price was only displayed for some hotel listings.
10 The primary judge found (against Trivago’s contention) that most consumers would only stay on the Trivago website for a short time. If a consumer clicked an Online Booking Site’s offer on the Initial Search Results Page, he or she would be referred away from the Trivago website to the Online Booking Site’s website in order to view that offer in greater detail and to complete a booking, if the consumer wished to do so.
11 An important aspect of the litigation was that the contractual terms between Trivago and the Online Booking Sites required the latter to pay Trivago a fee if a consumer clicked on the advertiser’s offer on Trivago’s website. This fee was described as the CPC or “cost per click”. The CPC would be payable regardless of whether the Online Booking Site converted the click into a booking. The CPC payable by the Online Booking Site, was the amount bid by them directly to Trivago. Trivago did not charge a fee to consumers for use of its website. Another issue in the litigation was whether consumers were informed about the existence or effect of the CPC, or whether or not they should have been.
12 The order of the Top Position Offer and the Second, Third and Fourth Position Offers was determined by Trivago’s proprietary Algorithm. Expert evidence was adduced as to various aspects of the Trivago Algorithm. The experts agreed that there were several inputs into the Algorithm, including the offer price, the historical click through rate for the hotel, the CPC, and the priority modifier.
13 Another issue at trial was the degree to which the CPC influenced the determination of the Top Position Offer. The respondent (the ACCC) particularly emphasised that most of Trivago’s revenue was derived from the CPC fees. Online Booking Sites made CPC “bids” to Trivago in respect of the offers they wished to have displayed when consumers conducted searches via the Trivago website.
14 The Online Booking Sites knew, because Trivago told them, that the higher a CPC bid, the more likely it was that an offer would be prominently displayed when a consumer conducted a relevant search. Whether Trivago adequately disclosed to consumers the role that the CPC bids played in determining the hotel offers displayed is one of the issues upon which this appeal turns.
15 Evidence was adduced at trial, not only as to the operation of the Trivago website at various times, but also as to Trivago’s advertising. Consumers of Trivago’s website were, by Trivago’s advertising, invited to look for their “ideal hotel for the best price” or “the best deal on accommodation”.
16 The ACCC contended that ordinary and reasonable users of Trivago’s website were looking for the best available hotel room offer that corresponded to their needs in relation to matters such as location, date and type of room at the cheapest available rate. It contended that the best available offer for the consumer’s chosen hotel was the cheapest offer that conformed to those needs, or an offer other than the cheapest offer if it had some other characteristic making it more attractive than cheaper available offers.
17 By making CPC bids, Online Booking Sites would compete to be displayed as the Top Position Offer which was displayed much more prominently and attractively by Trivago than other offers.
18 In the period from 1 December 2016 to 3 January 2018, there were over 20 million site visits where consumers clicked on Top Position Offers, but only 4.73 million visits where they clicked the More Deals slide-out showing offers additional to the Top Position Offer and the Second, Third and Fourth Position Offers shown on the Initial Search Results Page.
19 However, the Top Position Offer was the cheapest available offer for a hotel room only 33.2% of the time. The ACCC stressed that in 66.8% of cases, cheaper offers for the same room were available. Some cheaper available offers were not displayed at all because Trivago did not display offers from Online Booking Sites if the site’s CPC bid for the offer was below a minimum amount determined by Trivago.
20 The ACCC also asserted that when the Top Position Offer was more expensive than other available offers, this was not because there was some other characteristic that made the Top Position Offer more attractive to a consumer.
21 Rather, the ACCC said, amongst other things, that the most important factor that determined why a higher priced offer was displayed as the Top Position Offer above a lower priced offer for the same hotel was the value of the CPC bid by the competing Online Booking Sites.
22 During that part of the period when the Red Price was displayed above the Top Position Offer, the ACCC contended that this display implied a “like-for-like” comparison between available offers for the same room but without informing consumers clearly or at all that the Red Price might be higher because it related to a superior room or a room with superior amenity in the same hotel.
23 At trial, Trivago did not call any of its officers or employees to give evidence.
24 The findings against Trivago at trial were extensive. Some contraventions were admitted by Trivago and some contraventions, which were found by the primary judge to be established, are not the subject of appeal.
25 The facts recorded in this section were not in dispute.
3.1 Trivago’s advertising and marketing
26 During the Relevant Period, Trivago used a number of advertising methods to advertise its website to Australian consumers, including a series of television advertising campaigns.
27 Although there were numerous iterations and variations, each television advertisement adhered to a similar format in which a presenter would make oral statements about the Trivago website.
28 For example, in one of the advertisements that aired on Australian television networks between 10 January 2016 and 26 September 2017, the presenter (pictured below) stated as follows:
Have you ever looked for a hotel online?
Did you notice that there are so many different prices out there for the exact same room?
Trivago does the work for you and instantly compares the prices of over 600,000 hotels from over 200 different websites.
So, instead of searching for hours or spending too much, Trivago makes it easy for you to find the ideal hotel for the best price.
Just go to Trivago, type in where you want to go, and with two clicks select your check-in and check-out dates and search. It’s that simple.
Trivago searches hundreds of websites at the same time and shows you the most popular hotels. You can adjust the price so that it fits in with your budget, or select the number of stars, or filter by average guest ratings from over 100 million reviews.
Remember, Trivago shows you all the different prices for the exact same room, and that’s how you can be sure that you find your ideal hotel for the best price.
Hotel? Trivago
29 In addition, Trivago aired a number of advertisements throughout different times in the Relevant Period, in some cases starting before, but then continuing into the Relevant Period. Among those advertisements were the following that included statements relevant to the present appeal:
(a) between 28 August 2016 and 29 May 2017, Trivago caused a television advertisement to be aired on television networks in Australia stating:
there are so many different prices all over the internet. And Trivago instantly compares them all to find your ideal hotel for the best price
(b) from 13 January 2017 to 29 May 2017, Trivago caused television advertisements to be aired on television networks in Australia containing the following statements:
Trivago makes it easy for you to find the ideal hotel for the best price …
You can be sure that you can find your ideal hotel at the best price …
Remember, Trivago shows you all the different prices for the exact same room. And that’s how you can be sure that you find your ideal hotel for the best price
(c) between 20 December 2016 and 2 July 2018, Trivago caused television advertisements to be aired, which included the following statement:
and that’s one more way Trivago helps you find your ideal hotel at the best price.
30 In the period 29 June 2017 to 1 April 2018, a television advertisement with the following statements was aired on Australian television networks:
Did you notice that there are so many different prices out there for the exact same room?
I know I’ve been asking this a lot lately, but this is what it really means. The same hotel room can have up to ten different prices across the internet. So even if you spend a lot of time looking around, chances are, it is still out there for a better price. So make one last check before you book, the Trivago check.
Trivago compares prices from more than 200 websites to make sure you find your ideal hotel for the best price.
Hotel? Trivago
31 In the period between 19 December 2017 and 1 April 2018, a television advertisement with the following statements was aired on Australian television networks:
There are plenty of booking websites. Your ideal hotel may only be listed on one, or on several, but for very different prices. Do you really want to check them all?
Instead, use Trivago to compare these offers and find your ideal hotel for the best price.
Hotel? Trivago
32 In addition to the television advertisements, between 1 December 2016 and on or around 19 June 2017, Trivago caused the following statements to be made in a “snippet” which appeared beneath Google search results displayed when Australian consumers conducted a search for “Trivago”:
Compare over 250 booking sites and find the ideal hotel at the best price!
Compare hotels, find the cheapest price and guarantee the best deal on accommodation …
33 The Landing Page for the Trivago website displays the Trivago logo. During the first part of the Relevant Period (defined below), the logo was accompanied by the slogan: “Find your ideal hotel for the best price”. Later in the Relevant Period, this was changed to read: “Find your ideal hotel and compare prices from many websites”. It was subsequently changed again, by replacing the word “many” with the word “different”.
34 Because of various changes in the presentation of material, either advertising or content on the Trivago website, it was necessary to break up the Relevant Period in respect of the allegations raised by the ACCC into four periods, which the primary judge designated as follows (at [17]):
…
(a) the period from 1 December 2016 to 29 April 2018 (the first relevant sub-period), a period of approximately 17 months;
(b) the period from 29 April 2018 to 20 November 2018 (the second relevant sub-period), a period of approximately seven months;
(c) the period from 20 November 2018 to 13 February 2019 (the third relevant sub-period), a period of approximately three months; and
(d) the period from 13 February 2019 to 13 September 2019 (the fourth relevant sub-period), a period of approximately seven months.
35 It was then necessary to apply different descriptions to those periods, which his Honour dealt with in the following way (at [18]):
Consistently with the approach taken by the parties, the website during the first relevant sub-period will be referred to as website version 1, the website during the second relevant sub-period will be referred to as website version 2, the website during the third relevant sub-period will be referred to as website version 3 and the website during the fourth relevant sub-period will be referred to as website version 4. It should be noted that the appearance of the website was not necessarily static within each sub-period; on some occasions, as detailed below, changes were made within a sub-period.
3.2 The first relevant sub-period (1 December 2016 to 29 April 2018)
36 In the period from about 1 December 2016 to 4 January 2018, the slogan beneath the logo on the Landing Page read: “Find your ideal hotel for the best price”. Further below, under the heading “Find cheap hotels on trivago”, appeared the statement:
With Trivago you can easily find your hotel at the lowest rate. Simply enter where you want to go and your desired travel dates and let our hotel search engine compare accommodation prices for you.
(Emphasis in original.)
37 As at 4 January 2018, the top part of the Landing Page appeared as follows:
38 Prior to 12 April 2018, another slogan appeared on the Trivago website. This read: “Find your ideal hotel and compare prices from many websites”. From on or about 12 April 2018, the slogan was amended to state: “Find your ideal hotel and compare prices from different websites”.
39 The statement of agreed facts in the proceedings below included various screenshots of website version 1. These included the following screenshots of individual hotel listings:
___________________________________________________________________________
___________________________________________________________________________
___________________________________________________________________________
___________________________________________________________________________
___________________________________________________________________________
40 In response to a consumer search, website version 1 generated an Initial Search Results Page containing a number of individual hotel listings, each of which displayed:
(a) the Top Position Offer in the column furthest to the right. The Top Position Offer was the most prominently displayed offer for the hotel. It was displayed in large green text together with the Online Booking Site’s name and a green “view deal” click-out button. In addition, some hotel listings on the Initial Search Results Page displayed a box underneath the Top Position Offer showing that the Top Position Offer had particular attributes including, for example, “Pay at the hotel” or “Free cancellation” (see, for example, Fig 3 above);
(b) the Strike-Through Price in red strike-through text above the Top Position Offer;
(c) the Second, Third and Fourth Position Offers in smaller size text (compared with the Top Position Offer and Strike-Through Price). These offers were displayed less prominently than the Top Position Offer and were contained in one column;
(d) the More Deals button, with bolded black text indicating (until about 5 December 2017) the number of further offers contained in the More Deals slide-out (e.g. “More deals: 22”; see Fig 1 above);
(e) a red box above the Top Position Offer and Strike-Through Price which showed the percentage difference between the two prices (Percentage Savings box) (see Fig 2 above); and
(f) a Top Deal box would replace the Percentage Savings box if, among other factors, the Top Position Offer was at least 20% lower than the Strike-Through Price, the hotel had a rating of at least three stars and the hotel met Trivago’s internal rating index for quality (see Figs 1, 3 and 4 above).
41 Between May 2017 and January 2018, the Trivago website displayed the Strike-Through Price as the Fourth Position Offer (in addition to it appearing directly above the Top Position Offer; see Fig 1 and Fig 2). In or around January 2018, the Strike-Through Price ceased to be shown as the Fourth Position Offer (see Fig 3 and Fig 4).
42 In addition, during the first relevant sub-period, the following changes were made to the Initial Search Results Page:
(a) on or about 5 December 2017, the More Deals button on Trivago’s desktop website ceased to include the text indicating the numbers of further offers contained within the More Deals slide-out; instead, the bolded black text was changed to read “More deals from AU$...” and displayed the price of the cheapest deal contained in the More Deals slide-out (e.g. “More deals from AU$226”; see Figs 2, 3 and 4);
(b) on or about 6 October 2017, the Trivago website began displaying the Our Recommendations information button with a hover-over. If a consumer’s mouse cursor hovered over the Our Recommendations information button, the hover-over would display text stating:
Trivago gives you the opportunity to compare different offers for hotel rooms that are available through Trivago’s website. In some cases, lower prices might be available from other sources. In determining the price to display in the leading position of our search results, we consider a variety of factors, including price, the likelihood that you will find your ideal hotel, your ability to complete a booking after you click on a search result and the level of compensation provided by the booking sites we cover. In order to make more information available on pricing options to our users, additional prices are listed in the “More deals” slide-out.
(see the “Mantra Macarthur” listing above.)
(c) from about 27 January 2018, the text displayed if a consumer’s mouse cursor hovered over the Our Recommendations information button stated:
The ranking results reflect your search criteria and our assessment of the attractiveness of the offer compared to other offers available on our site. It also reflects the compensation paid by the booking site.
(d) on or about 29 March 2018, a hover-over was added in respect of the Strike-Through Price. If a consumer’s mouse cursor hovered over the Strike-Through Price, the following text was displayed:
The strike-through price corresponds to the cheapest offer we received from the most expensive booking site on Trivago for this hotel and your stay dates.
(see the “Four Seasons Hotel” listing above).
(e) on or about 29 March 2018, a hover-over was added for the Percentage Savings box. If a consumer’s mouse cursor hovered over the Percentage Savings box, the following text was displayed:
This figure represents the difference between the strike-though deal price and the featured deal price [i.e. the Top Position Offer].
(see Fig 4 above).
(f) on or about 29 March 2018, a hover-over was added for the Top Deal box (which was only displayed if the conditions were met). If a consumer’s mouse cursor hovered over the Top Deal box, the following text was displayed:
This deal is at least 20% cheaper than the strike-through deal price and its Trivago Rating Index is above 7.5.
3.3 The second relevant sub-period (29 April 2018 to 20 November 2018)
43 From 12 April 2018 (i.e. just before the start of the second relevant sub-period), the words on the Landing Page read: “Find your ideal hotel and compare prices from different websites”. During the second, third and fourth relevant sub-periods, the Landing Page therefore appeared as follows:
44 In response to a consumer search, the Initial Search Results Page in website version 2 displayed the same features as website version 1 (set out at [40]-[42] above).
45 However, during the second relevant sub-period, the following changes to the Initial Search Results Page occurred:
(a) in or around August 2018, the text displayed if a consumer’s mouse cursor hovered over the Strike-Through Price was amended to state:
This is the cheapest deal from the most expensive booking site with offers for this hotel on your stay dates.
(b) on or about 6 August 2018, the Top Deal box was removed from the Trivago website.
46 Website version 2 also had the following new features:
(a) from about 3 July 2018, some information about Trivago’s Algorithm and the ranking of offers on the Initial Search Results Page was provided to consumers under the link “Learn how Trivago works”; and
(b) in about October 2018, the text displayed if a consumer’s mouse cursor hovered over the Our Recommendations information button was amended to include a hyperlink to the “Learn how Trivago works” page. Consumers who clicked on that link were taken to a new page on the Trivago website that stated:
How does Trivago determine the ‘our recommendations’ sort?
The ‘our recommendations’ feature is based on a dynamic Algorithm that shows you a range of attractive and relevant offers we think you’re going to love. In the ‘top position’ we display in green the offer which our Algorithm recommends as a great offer. Our Algorithm takes into account a number of relevant factors, such as your search criteria (for example your location and stay dates), the offer’s price, and its general attractiveness – for example the experience we think you’ll likely have on the displayed booking site. We also take into account the compensation booking sites provide us with when a user clicks on an offer.
To the left of the ‘top position’ offer and/or under the ‘more deals from’ tab, we display additional offers to allow you to compare other deals for the same hotel. You may find offers under the ‘more deals from’ slide-out section with a lower price than the ‘top position’ offer. While price is an important factor when selecting our ‘top position’ offer, we believe other factors, such as those mentioned above, make offers attractive and relevant to you, and contribute to high levels of satisfaction by our users.
Ultimately, our mission is to give you the information and tools to help you find your ideal hotel.
3.4 The third relevant sub-period (20 November 2018 to 13 February 2019)
47 Website version 3 was displayed during this period.
48 In response to a consumer search, the Initial Search Results Page in website version 3 displayed:
(a) the Top Position Offer in the column furthest to the right. The Top Position Offer was the most prominently displayed offer on the Trivago website. It was displayed together with the Online Booking Site’s name and a green “view deal” click-out button. In addition, some hotel listings on the Initial Search Results Page displayed a box underneath the Top Position Offer showing that the Top Position Offer had particular attributes including, for example, “Pay at the hotel” or “Free breakfast” (consistently with website versions 1 and 2);
(b) (for some of the hotel listings only) a Red Price from identified Online Booking Sites in smaller text above the Top Position Offer with a hover-over. If a consumer’s mouse cursor hovered over the Red Price, the following text was displayed:
For your stay dates, this is the cheapest deal from the most expensive booking site with offers for this hotel. It may relate to a room type or amenities package that differs from other offers on our site.
(see “Punthill Northbank” listing below).
(c) the Second, Third and Fourth Position Offers in smaller size text (compared with the Top Position Offer) to the left of the Top Position Offer. These offers were displayed less prominently than the Top Position Offer and were contained in one column (consistently with website versions 1 and 2);
(d) the More Deals button as described in [42(a)] above; and
(e) the Our Recommendations information button as set out in [42(c)] and [46(b)] above.
49 Website version 3 did not include the following features:
(a) the Strike-Through Price (save to the extent the Strike-Through Price continued to appear in the map portion of the Trivago website);
(b) the Percentage Savings box; and
(c) the Top Deal box.
50 On or about 14 January 2019, it came to Trivago’s attention that the words “best price” appeared in a subset of unpaid Trivago-related Google search results in Australia. To produce its unpaid search results, Google ‘crawls’ internet pages, including the metadata of those pages. Although Trivago updated the metadata of the Trivago website, Trivago inadvertently did not update the metadata of the pages for individual hotels on the Trivago website. As a result, Google search results for individual hotels (e.g., for a Trivago listing for the Hilton in Sydney) included a snippet that included the words “best price”. Trivago subsequently updated the metadata for these pages. This error did not affect Trivago’s paid advertisements or its other unpaid listings on Google.
3.5 The fourth relevant sub-period (13 February to 13 September 2019)
51 On or about 13 February 2019, it came to Trivago’s attention that Strike-Through Prices continued to appear on the map portion of the Trivago website. When searching for hotels in a particular location, a user could select an option to view a map that included relevant offers. The map was inadvertently not updated and, as a result, Strike-Through Prices continued to be displayed when users hovered over price symbols on the map. Trivago removed these Strike-Through Prices promptly after becoming aware of this issue on or about 13 February 2019. Therefore, during the fourth relevant sub-period, the Strike-Through Price no longer appeared in the map portion of the Trivago website.
52 Website version 4 displayed the following:
53 In response to a consumer search, the Initial Search Results Page for website version 4 displayed the same features as website version 3 (as described at [48] above.
54 From about 17 April 2019 the words “How payments to us affect ranking” appeared next to the Our Recommendations information button on website version 4. The hover-over accompanying the Our Recommendations information button continued to read as set out in [42(c)] and [46(b)] above.
55 Trivago informed the ACCC by correspondence of the following matters, which were treated as proven facts:
(a) in the period 1 December 2016 to 3 January 2018 new search results were displayed on Trivago’s Australian website 217,514,783 times;
(b) Trivago does not record the number of users visiting its Australian website or the number of times a user enters a location or region in Trivago’s search bar. However, Trivago records the number of user “sessions”. A session starts when a user first interacts with the Trivago website and continues until the user has been inactive for one hour. For example, if a user visits Trivago’s Australian website and is then inactive for an hour before interacting with the site again, two user sessions will be logged by Trivago’s system;
(c) in the period 1 December 2016 to 3 January 2018 there were 4,421,184 sessions on Trivago’s Australian website where the sorting order was changed from the default sorting option to a different sorting option;
(d) in the period 1 December 2016 to 3 January 2018 there were 4,726,241 sessions on Trivago’s Australian website where the More Deals slide-out was displayed;
(e) in the period 1 December 2016 to 3 January 2018 there were 1,400,205 sessions on Trivago’s Australian website where a user clicked on a deal displayed in the More Deals slide-out;
(f) in the period 1 December 2016 to 3 January 2018 there were 20,039,530 sessions on Trivago’s Australian website where the Top Position offer for any hotel listing was clicked; and
(g) in the period 1 December 2016 to 3 January 2018 there were 68,783 sessions on Trivago’s Australian website where a comparison rate, being the Fourth Position Offer that was the same as the Strike-Through Price (see Fig 2 above) was clicked.
4. THE ASSERTED REPRESENTATIONS
56 The ACCC contended that Trivago made a number of key representations throughout the Relevant Period (with some representations made only during certain sub-periods) that amounted to conduct in contravention of s 18 (misleading or deceptive conduct), s 29(1)(i) (false or misleading representations about the price of goods and services) and s 34 (misleading conduct as to the nature of services) of the Australian Consumer Law (sch 2 to the Competition and Consumer Act 2010 (Cth)). It sought declarations to that effect.
57 Those asserted representations were as follows:
(a) The Cheapest Price Representation during the first relevant sub-period. Through online and television advertising, Trivago is alleged to have represented that its website would quickly and easily identify the cheapest rates available for a hotel room responding to a consumer’s search when in fact the Trivago website did not enable consumers to quickly or easily identify the cheapest rates available for particular hotel rooms. This is said to contravene s 18 and s 34 of the Australian Consumer Law;
(b) The Top Position Representation during the entire Relevant Period, which is said to contravene s 18 and s 29(1)(i) of the Australian Consumer Law by prominently displaying on the Trivago website, Top Position Offers (which were sometimes accompanied by a Top Deals box or a Percentage Savings box) and thereby representing that the Top Position Offers were the cheapest available offers for an identified hotel, or had some other characteristic which made them more attractive than any other offer for that hotel when in fact:
(i) the Top Position Offers were not always the cheapest available offers for an identified hotel;
(ii) Trivago did not select the Top Position Offers by reference to any other characteristic which may have made them the most attractive offer; and
(iii) Trivago selected the Top Position Offers primarily by reference to the CPC bids Trivago would receive from the Online Booking Site if a consumer clicked on that Online Booking Site’s offer.
(Emphasis added.)
(c) The Strike-Through Representation during the first and second relevant sub-periods, which is said to contravene s 18 and s 29(1)(i) of the Australian Consumer Law by displaying the Strike-Through Price above the Top Position Offer on the hotel listing display and the map portion of the Trivago website, and thereby representing that the Strike-Through Price was a comparison between prices offered for the same room category in the same hotel when in fact the Strike-Through Price that Trivago displayed often related to a more expensive room category than the Top Position Offer;
(d) The Red Price Representation, which together with the Strike-Through Representation during the third relevant sub-period is said to contravene s 18 and s 29(1)(i) of the Australian Consumer Law by:
(i) displaying the Red Price above the Top Position Offer on the hotel listing display, and thereby representing that the Red Price was a comparison between prices offered for the same room category in the same hotel when in fact the Red Price that Trivago displayed often related to a more expensive room category than the Top Position Offer; and
(ii) displaying the Strike-Through Price above the Top Position Offer on the map portion of the Trivago website, and thereby representing that the Strike-Through Price was a comparison between prices offered for the same room category in the same hotel, when in fact the Strike-Through Price that Trivago displayed often related to a more expensive room category than the Top Position Offer.
58 Further, during the fourth relevant sub-period, when the Strike-Through Price was completely removed from the Trivago website, the ACCC contends that Trivago continued to make the Red Price Representation by engaging in the same conduct as described at (d)(i) above.
59 The ACCC also sought declarations in relation to the additional conduct allegations. In relation to the first relevant sub-period, it was alleged that Trivago contravened s 18 and s 34 of the Australian Consumer Law by:
…
making the Cheapest Price Representation, and by making Top Position Offers together with the Top Position Representations and the Strike-Through Representation, and thereby leading consumers to believe that the Trivago website provided an impartial, objective and transparent price comparison which would enable them to quickly and easily identify the cheapest available offer for a particular (or the exact same) room at a particular hotel when, in fact, the Trivago website:
(c) did not enable consumers to quickly or easily identify the cheapest prices available for a particular hotel room; and
(d) directed consumers to more prominently displayed offers made by Online Booking Sites that offered to make higher CPC payments to Trivago than competing Online Booking Sites each time a consumer clicked on the offer, irrespective of the prices offered to consumers for a particular hotel room.
60 In relation to the second and third relevant sub-periods, the ACCC sought a declaration in similar terms, save that, rather than relying on the Cheapest Price Representation (which is alleged only in respect of the first relevant sub-period), the ACCC relied on Trivago making Top Position Offers on the Trivago website, making the Top Position Representation, making the Strike-Through Representation, and:
advertising the Trivago website using statements such as: “Compare over 250 booking sites and find the ideal hotel at the best price!”; “find your ideal hotel and compare prices from different websites”; “impartial comparison”; “find the ideal hotel at a great price”; and “Compare & Save. No Ads or Pop-Ups”.
(Emphasis added.)
61 In relation to the fourth relevant sub-period, the declaration sought was similar to that sought in respect of the second and third relevant sub-periods. The ACCC relied on Trivago making Top Position Offers on the Trivago website, making the Top Position Representation, making the Red Price Representation, and advertising the Trivago website using statements such as: “Find your ideal hotel and compare prices from different websites”; “impartial comparison”; “find the ideal hotel at a great price”; and “Compare & Save. No Ads or Pop-Ups”.
62 The ACCC also contended in its Further Amended Concise Statement (FACS) (at [12]-[19]) that:
12 Contrary to the Top Position Representation, the Top Position Offer was not always the cheapest available offer and Trivago did not select it by reference to any other characteristic which may have made it the most attractive offer. Trivago selected the Top Position Offer primarily by reference to the value of the CPC Trivago would receive from the advertiser who submitted the offer.
13 Trivago used an Algorithm to determine the Top Position Offer. The Algorithm placed a significant weighting on the value of an advertiser’s CPC bid. By applying the Algorithm, Trivago gave greater prominence to offers from advertisers who had submitted higher CPC bids than to cheaper offers at the same hotel from advertisers who submitted lower CPC bids. In addition, from at least May 2017 to at least October 2017, Trivago set a minimum CPC amount. If an advertiser submitted a CPC bid below this amount, the Algorithm prevented the Trivago website from displaying the advertiser’s offer, even if it was the cheapest price offer Trivago had received. The result was that often the Top Position Offer was not the cheapest price offer Trivago had received for the relevant hotel.
…
15 Beneath the Fourth Position [Offer], Trivago displayed a “More Deals” button. Clicking on this button opened a list of other offers which advertisers had submitted for the same hotel. From at least May 2017 to at least October 2017, the “More Deals” button did not display the cheapest price offer which Trivago had received for that hotel. In addition (as set out in paragraph 13) if an advertiser made a CPC bid that was less than a predetermined minimum, Trivago did not display the advertiser’s offer on its website at all, even if it was the cheapest price offer Trivago had received for the hotel. From a date not before October 2017, Trivago changed the “More Deals” button to display the cheapest offer it had received and introduced a mouse over information button in respect of the top position Algorithm. These changes did not affect the Top Position Representation.
…
19 By making the Strike-Through Representation and the Red Price Representation (separately or in combination) [and] by displaying the Percentage Saving or Top Deal box next to the Top-Position Offer, Trivago reiterated and reinforced the Top Position Representation. One effect of this was that consumers (especially those who had viewed advertising in which Trivago made the Cheapest Price Representation) were less likely to open the “More Deals” list, even after Trivago amended the button to identify that cheaper price offers were available. …
(Emphasis added.)
63 In a confidential annexure to the FACS, the ACCC compared the number of visits to the Trivago website in which a consumer clicked on an offer located in the More Deals slide-out with the number of visits in which a consumer clicked on a Top Position Offer. The figures provided related to the period from 1 December 2016 to 3 January 2018, which captures most of the first relevant sub-period. During this period the latter figure was much higher than the former.
64 By its Concise Response, Trivago raised various matters, including the following:
5. Trivago acts as a metasearch site and not as a direct booking site. Trivago’s primary function is to aggregate and compare accommodation offers from different Online Booking Sites in a manner that consumers find helpful. The booking sites appearing on Trivago’s website include participating hotels and online travel agencies. If a consumer clicks on the booking site’s offer on the Trivago website, the consumer is taken to the booking site’s online booking service and completes the booking using the booking site’s service. Trivago does not provide the booking service.
6. Trivago’s contractual terms require booking sites to pay Trivago a fee if a consumer clicks on the booking site’s offer on the Trivago website (referred to as the “cost-per-click” or “CPC”). The CPC is Trivago’s principal source of revenue. Trivago does not charge fees to consumers who use its service.
7. Trivago’s business offers booking sites a marketplace on which to advertise their hotel offers. The CPC payable by the booking site to Trivago to advertise their hotel offer is the amount bid by the booking site on Trivago’s marketplace, subject to a minimum CPC determined by Trivago.
(Emphasis added.)
…
C. Lowest rate statements
24. Trivago admits that during the following periods, it described and marketed its Australian website in the following mediums and using statements to the following effect:
(a) between 1 December 2016 and 4 January 2018, as alleged by the ACCC, the Trivago website stated:
“Find your ideal hotel for the best price”;
and
“With Trivago you can easily find your ideal hotel at the lowest rate. Simply enter where you want to go and your desired travel dates and let our hotel search engine compare accommodation prices for you.”;
(b) between 1 December 2016 and 19 June 2017, as alleged by the ACCC, Trivago caused the following statements to be made in the snippet which appeared beneath the Google search results displayed when Australian consumers conducted an online search for Trivago:
“Compare over 250 booking sites and find the ideal hotel at the best price!”;
and
“Compare hotels, find the cheapest price and guarantee the best deal on accommodation...”;
(c) between 28 August 2016 and 29 May 2017, as alleged by the ACCC, Trivago caused a television advertisement to be aired on television networks in Australia on 10,933 occasions stating:
“...there are so many different prices all over the internet. And Trivago instantly compares them all to find your ideal hotel for the best price.”;
(d) from 13 January 2017 to 29 May 2017, as alleged by the ACCC, Trivago caused television advertisements to be aired on television networks in Australia on 4,761 occasions containing the following statements:
“Trivago makes it easy for you to find the ideal hotel for the best price”;
and
“You can be sure that you can find your ideal hotel at the best price”;
and
“Remember, Trivago shows you all the different prices for the exact same room. And that’s how you can be sure that you find your ideal hotel for the best price.”
25. Trivago admits that the statements referred to in paragraph 24 (lowest rate statements) conveyed to ordinary consumers that:
(a) the Trivago website is easy to use; and
(b) the Trivago website assists consumers in finding their ideal hotel accommodation at the lowest rates advertised by a wide range of booking sites through the Trivago website for that hotel on the relevant stay dates.
26. Trivago says that, at all relevant times:
(a) the Trivago website has been easy to use; and
(b) the Trivago website has in fact assisted consumers in finding their ideal hotel accommodation at the lowest rates advertised by a wide range of booking sites through the Trivago website for that hotel on the relevant stay dates.
27. However, for the purposes of this proceeding, Trivago admits that:
(a) while the Initial Search Page Offers were selected by Trivago from the lowest rates advertised by those booking sites on the Trivago website for that hotel on the relevant stay dates, those rates were not always the lowest rates advertised on the Trivago website for that hotel on the relevant stay dates;
(b) the lowest rate statements may have caused some consumers to form an erroneous belief that the Initial Search Page Offers were the lowest rates advertised on the Trivago website for that hotel on the relevant stay dates; and
(c) to that extent, during the relevant period until 29 April 2018, Trivago engaged in conduct in contravention of s 18 and s 34 of the Australian Consumer Law (ACL).
28. As noted above, from on or about 29 April 2018, Trivago ceased making the lowest rate statements.
D Strike through pricing
29. Trivago admits that, during the relevant period until on or about 20 November 2018, in response to a consumer search for accommodation in a particular region on particular stay dates, the initial search results page on Trivago’s website displayed, in respect of each hotel listed:
(a) a Strike-Through Price with a “hover-over” which stated: “This is the cheapest deal from the most expensive booking site with offers for this hotel on your stay dates.”; and
(b) in the circumstances referred to earlier, a Percentage Savings box and a Top Deal box.
30. For the purposes of this proceeding, Trivago admits that:
(a) while the Strike-Through Price for each hotel listed on the initial search results page was the lowest rate from the most expensive booking site with offers for that hotel on the relevant stay dates, the room rate did not always relate to the same room category as the Top Position Offer;
(b) by displaying the Strike-Through Price next to the Top Position Offer in the form it was displayed (using a “strike-through” notation) either on its own or in conjunction with the Percentage Savings box and the Top Deal box, Trivago may have caused some consumers to form an erroneous belief that the Top Position Offer and the Strike-Through Price were offers for rooms in the same room category; and
(c) to that extent, by displaying the Strike-Through Price next to the Top Position Offer in the form it was displayed during the relevant period until on or about 20 November 2018 either on its own or in conjunction with the Percentage Savings box and the Top Deal box, Trivago engaged in conduct in contravention of s 18 and s 29(1)(i) of the Australian Consumer Law (ACL).
31. As noted above, from on or about 20 November 2018, Trivago altered the “hover-over” in respect of the Strike-Through Price to state that the price may relate to a room type or amenities package that differs from other offers on the site.
32. Further, from on or about 20 November 2018, Trivago ceased using the “strike-through” notation and also removed the Percentage Savings box and the Top Deals box.
5. THE CONTRIBUTION OF EXPERT EVIDENCE
65 There was expert evidence at trial. It was relevant not to the question of whether representations were made, a question for the primary judge, but to the question of the underlying content and computation of data on the website and to the habits of relevant consumers, being a topic going to their likely perceptions on using the Trivago website. The expert evidence was also relevant to the question of the accuracy of the content on the Trivago website.
66 The ACCC called two witnesses:
Professor Robert Slonim, a Professor in the School of Economics, Faculty of Arts and Social Sciences at the University of Sydney; and
Mr Victor Bajanov, who holds the position of Executive, Product Analytics at Quantium, a global data and analytics firm headquartered in Sydney.
67 Trivago also called two expert witnesses:
Professor David C Parkes, a Professor in the School of Engineering and Applied Sciences at Harvard University; and
Mr Patrick Smith, partner at RBB Economics.
68 Additionally, there was a joint report dated 22 August 2019 of Mr Bajanov and Professor Parkes, which highlighted their areas of agreement and disagreement (the Joint Report). They gave evidence concurrently.
69 In relation to the computer science experts, Mr Bajanov and Professor Parkes, the primary judge was satisfied that each expressed opinions independently and impartially and each was knowledgeable about the topics they were questioned on. There was an initial report from Mr Bajanov dated 15 May 2019 (the First Bajanov Report) addressing nine questions that had been set out in his letter of instructions. Professor Parkes responded by a report dated 15 July 2019 (the First Parkes Report). Mr Bajanov prepared a reply report on 5 August 2019 (the Second Bajanov Report). Professor Parkes prepared a further report on 22 August 2019 (the Second Parkes Report). Their concurrent evidence was structured around the nine questions set out in the Joint Report. It was conducted in closed court due to the commercial sensitivities of some aspects of Trivago’s Algorithm and other data provided by Trivago.
70 In considering the evidence given by these experts by reference to the nine common questions, the primary judge recorded the following evidence and his findings:
5.1.1 Question 1: Making reference to the source code, pseudocode and written descriptions in the documents provided, provide a brief, high-level description of how the Top Position and Default Sorting Algorithms operated
71 The primary judge found there was general agreement between the experts in relation to this question. Trivago’s Algorithm calculated a “composite score” for each offer made by an Online Booking Site with respect to a hotel listing. (The word “offer” referred to the offer price for the hotel room, rather than the CPC bid.) The offer with the highest composite score became the Top Position Offer for that hotel listing.
72 The inputs to the composite score (as set out in the Joint Report) were:
(a) the offer price, the CPC and the priority modifier;
(b) hotel position, average historical price, historical click-through rate; and
(c) minimum listing price, maximum listing price, minimum priority CPC, maximum priority CPC, minimum gain and maximum gain.
73 The priority modifier was a value that adjusts the CPC value, as discussed further below. The priority CPC was the CPC multiplied by the priority modifier. The expression “gain” was a quantity related to the estimated revenue Trivago could expect to receive from showing an offer, calculated as the product of the predicted click-through rate on the offer and the priority CPC.
74 The priority modifier, described by Professor Parkes as a “kind of quality adjustment”, was made to the CPC bid of an Online Booking Site. While there was agreement between the experts that the priority modifier was strongly correlated with, or heavily dependent on, a value called the “Landing Page score”, the evidence regarding the Landing Page score was incomplete. Professor Parkes’ understanding of the Landing Page score was based on the information that Trivago provided to Online Booking Sites and Trivago’s Response to the ACCC, but the statements in those documents were not proved in evidence by Trivago. Trivago did not call any of its officers or employees to give evidence.
75 Thus, the primary judge said that while it may be accepted that Trivago stated in those documents that the Landing Page score is a measure of the “quality” of the Online Booking Site’s Landing Page in terms of user experience, he was not prepared to accept that the Landing Page score actually operated in that way. As Professor Parkes said in his evidence, neither he nor Mr Bajanov had been able to analyse how Trivago went about calculating the Landing Page score. Further, Professor Parkes made clear that his conclusions regarding the Landing Page score were “contingent” on the statements made by Trivago to the Online Booking Sites and to the ACCC. In any event and fundamentally, the primary judge noted that, even if the Landing Page score did operate in the manner understood by Professor Parkes, it measured the quality of an Online Booking Site’s landing page, not the quality of the hotel accommodation offers themselves.
5.1.2 Question 2: What factors influenced whether an offer was selected as the Top Position Offer?
76 The experts agreed in the Joint Report that the following factors were the raw inputs into Trivago’s Algorithm:
(a) The minimum listing price: the minimum price among the competing offers within a listing;
(b) The maximum listing price: the maximum price among the competing offers within a listing;
(c) The hotel position in the search results;
(d) The historical click through rate for the hotel;
(e) The average historical price of the hotel over a pre-defined period;
(f) The offer price: the price provided by the Online Booking Site for a specific offer within a listing;
(g) The CPC: the CPC bid made by an Online Booking Site for a specific offer;
(h) The priority modifier: see the discussion in relation to question 1, above;
(i) The minimum gain and maximum gain of all offers within the listing; and
(j) The minimum priority CPC and maximum priority CPC for all offers within the listing.
5.1.3 Question 3: What was the weighting or relative importance of these factors?
77 This was the main area of disagreement between the experts. In the First Bajanov Report, Mr Bajanov utilised two methods to consider question 3:
(a) first, he considered the effect of the various factors on the composite score by reverse engineering Trivago’s Algorithm; and
(b) secondly, he considered the effect of the various factors on an offer being selected as the Top Position Offer by building a ‘Gradient Boosting Model’ (explained below).
78 Mr Bajanov had been provided with Trivago data for three dates (9 December 2017, 3 January 2018 and 5 April 2018) in respect of four capital cities. The reason Mr Bajanov adopted a reverse engineering approach in his first method was because he had not, at that stage, been provided with certain ‘weights’ used in the Trivago Algorithm. However, the primary judge said that in terms of the disagreement between the experts, little, if anything, turned on the fact that Mr Bajanov used a reverse engineering approach. This was because the major area of disagreement between the experts concerned Mr Bajanov’s second method, rather than his first method. Applying his first method, Mr Bajanov concluded that the three factors with the most influence on the composite score on 9 December 2017 were (in order of importance):
(a) offer price;
(b) CPC; and
(c) minimum listing price.
79 The concept of a Gradient Boosting Model was explained as being a machine learning algorithm commonly used by data science professionals to predict a single outcome given a set of raw input data; it was widely used due to its high accuracy and its ability to determine the relative contributions of the input factors to each predicted value. Applying this method, Mr Bajanov concluded that the CPC had the highest relative importance in determining which offer was most likely to be selected as the Top Position Offer, followed by the maximum priority CPC and the priority modifier. For example, he concluded that on 9 December 2017, the factors had the following relative importance:
(a) CPC (59.0%);
(b) maximum priority CPC (21.0%); and
(c) priority modifier (9.6%).
80 Under the Gradient Boosting Model, the CPC was the most important factor on all three dates, with a relative importance of between 52.9% and 59.2%. The range for the priority modifier was from 8.7% to 18.3%. The range for the offer price was from 0.7% to 1.0%.
81 Professor Parkes strongly disagreed with the appropriateness of the Gradient Boosting Model utilised by Mr Bajanov. He based his analysis instead on Trivago’s Algorithm, in circumstances where he had access to the “weights” used in the Algorithm, which Mr Bajanov did not have.
82 Applying his methodology, Professor Parkes concluded that:
(a) the offer price was the most important factor in selecting the Top Position Offer, with a relative importance between 45.5% and 60.0%;
(b) the CPC was the second most important factor on all three dates, with a relative importance between 33.8% and 44.8%;
(c) the priority modifier was the third most important factor on all three dates, with a relative importance between 4.0% and 13.3%; and
(d) the other factors set out above (at [76]) have no bearing.
83 Ultimately, the primary judge concluded that it was not necessary to form a concluded view as to which of the experts’ approaches was preferable. Mr Bajanov’s analysis involved looking at the events that happened and assessing the relative importance of the various factors in the selection of a particular offer as the Top Position Offer. Professor Parkes’ analysis involved assessing the relative importance of the various factors by making adjustments to those factors and assessing the effect of making such adjustments on the selection of an offer as the Top Position Offer. Each method addressed a different question, and the answers to both questions were of assistance to the primary judge in resolving the issues at first instance; in particular, the question whether, if the alleged representations were made, they were misleading or deceptive or liable to mislead.
84 Most importantly, however, the primary judge noted that, even on Professor Parkes’ approach, the CPC was a very significant factor in determining the Top Position Offer: it was the second most important factor, with a relative importance of between 33.8% and 44.8%. This was directly relevant to the contention of the ACCC in [13] of its FACS, rather than in [12] of the FACS (set out at [62] above).
5.1.4 Question 4: How and to what extent (if any) did the Top Position offer Algorithm filter certain offers made by advertisers?
85 It was common ground between the experts that Trivago’s Algorithm filtered out offers which did not meet a minimum gain threshold. Offers were also filtered out by a minimum CPC bid, the value of which was confidential.
5.1.5 Question 5: How important was an advertiser’s CPC payment in determining the Top Position Offer?
86 There was disagreement between the experts in relation to this question, but it is covered under question 3 in 5.1.3. The experts also agreed on several matters in relation to this question, including (as set out in the Joint Report) that 84.52% of listings had a Top Position Offer with an offer price lower than the mean of all offer prices submitted by Online Booking Sites, and a CPC higher than the mean of all CPC bids. The mean CPC within a listing was highest when the offer prices submitted varied the least within a listing. This demonstrated that high CPC and low-priced offers were not necessarily inconsistent. This statement was subject to the qualification that Mr Bajanov could not verify the calculations for the figure of 84.52%.
87 Professor Parkes said:
(a) approximately 86.5% of listings had a Top Position Offer lower than the mean of all offers; and
(b) conversely, approximately 13.5% of listings had a Top Position Offer higher than the mean of all offers.
5.1.6 Question 6: How “frequently” were higher priced hotel offers selected as the Top Position Offer over alternative, lower priced offers? In these cases (if any), what factor primarily determined why a higher priced offer was selected over a lower-priced offer?
(a) higher priced hotel offers were selected as the Top Position Offer over alternative lower priced offers in 66.8% of listings;
(b) in regard to the comparison with the lowest priced offer, the price of the Top Position Offer was the same in 33.2% of listings; within 1% in 63.1%, within 5% in 79.3%, within 10% in 88.7%, within 15% in 95.3%, and within 20% in 97.3%; and
(c) when a higher priced offer was selected as the Top Position Offer over the lowest priced offer, then the observable variation between the two offers was explained by a difference in CPC more than any other factor. The next most common factor was the priority modifier, which was the leading factor in between 2% and 8% of listings, depending on city and date.
89 The primary judge noted (at [126]) that in oral evidence the experts confirmed that the 66.8% figure referred to above (at [88(a)]) was based on their analysis of all the data provided to them; in other words, it included both offers that were displayed on the Trivago website and offers that were not displayed (i.e. those that were filtered out). In relation to [88(b)] above, the effect of these figures was that, in approximately 88% of cases, the variation between the Top Position Offer and the cheapest offer was less than 10%. Conversely, in approximately 12% of cases, the variation between the Top Position Offer and the cheapest offer was more than 10%.
5.1.7 Question 7: What percentage of listings had a Top Position Offer with the highest CPC of all offers within that listing? Further, how did CPC generally relate to offer price?
90 The experts agreed that 56% of the listings had a Top Position Offer with the highest CPC.
5.1.8 Question 8: What percentage of listings had a Top Position Offer with the cheapest price offer of all offers within that listing? What percentage of listings had a Top Position Offer within 1% or 5% of the cheapest offer of all offers within that listing?
91 The experts agreed that:
(a) as noted in answer to question 6 above, 33.2% of listings had a Top Position Offer that was the cheapest offer; and
(b) when the Top Position Offer was 10% or 20% higher than the minimum listing price, there was a discontinuity in the distribution of the price of other offers at the same 10% or 20% price points. The experts said that they had not been provided with data to definitively establish the cause of the discontinuity.
The statement in (b) was subject to the qualification that Mr Bajanov did not verify the calculations.
5.1.9 Question 9: Considering only listings where the Top Position Offer was not the cheapest price offer of all offers within that listing, what percentage of the Top Position Offers had a CPC value higher than the cheapest offer? Further, what were the other factors by which the Top Position Offer and cheapest price offer differed?
92 The experts agreed that, considering offers where the Top Position Offer was not the cheapest offer of all offers within a listing, 95.9% of the Top Position Offers had a CPC value higher than that of the cheapest offer.
93 There was a difference of opinion between the experts in relation to the second part of question 9. Professor Parkes referred to economic principles and academic literature concerning consumer behaviour in relation to the attributes of a hotel room. He stated that the academic literature documented that consumers believe that offers for a given hotel room are heterogeneous. Professor Parkes stated that he had created a dataset of listings where the Top Position Offer differed by more than 1% from the lowest price offer and had examined the variations in attributes between the Top Position Offer and the lowest price offer for the listings in this dataset. His analysis did not attempt to quantify the value of those attributes.
94 As the primary judge explained:
136 … in order to create this dataset, he searched the Trivago website for a one-night stay in Adelaide, Melbourne, Perth and Sydney four days in advance of check-in. (This was seen as a typical search that a consumer would do on the Trivago website ...) The search was carried out on 11 July 2019. From these searches, Professor Parkes examined all first page listings for which the Top Position Offer was at least 1% more than the lowest price offer listed. This yielded 27 listings. For each of the corresponding pairs of Top Position Offer and lowest priced offer, Professor Parkes recorded whether each of the Top Position Offer and lowest price offer included the following attributes: free cancellation; bed size; free Wi-Fi; pay at hotel; and free breakfast. The results of this analysis were set out in the report (and are set out later in these reasons).
137 Following criticism of this analysis in the Second Bajanov Report, including on the basis of sample size, Professor Parkes supplemented his analysis by collecting additional datasets of listings on 9 August 2019 and 13 August 2019 ... Professor Parkes stated that this further analysis reaffirmed the conclusions in … his first report, namely that there is clear evidence of heterogeneity among offers for a hotel listing. Again, Professor Parkes examined all first page listings for which the Top Position Offer was at least 1% more than the lowest price offer listed. This yielded 28 listings for 9 August 2019 and 30 listings for 13 August 2019. For each of the corresponding pairs of Top Position Offer and lowest price offer, Professor Parkes carried out the same analysis as described above.
138 The results of this analysis were set out in Figures 6, 7 and 8 in the Second Parkes Report (including, at Fig 6, the results for 11 July 2019):
139 In the above figures, the statement “TPO [Top Position Offer] is equivalent to lowest price offer” means that the offering is equivalent in respect of the particular attribute – either they both have it or they both do not have it ... The statement “TPO is better than lowest price offer” means that the Top Position Offer is better than the lowest price offer in respect of the particular attribute – for example, the Top Position Offer has free Wi-Fi and the lowest price offer does not.
140 It should be noted that the above analyses were not based on the Trivago-provided dataset analysed by the experts and discussed above in relation to earlier questions. That dataset related to three earlier dates. Rather, the analyses set out in Figures 6, 7 and 8 were based on more recent searches carried out by Professor Parkes.
141 In oral evidence, Professor Parkes stated that: the results of these analyses were statistically significant; 87% of the time the Top Position Offer and the lowest price offer differed in non-price ways; moreover, where there was a statistically significant effect, it went in favour of the Top Position Offer ... Professor Parkes provided the following further observations, referring to his discussion with Mr Bajanov …:
And so where we have a discussion, I’m not even sure if it rises to a disagreement. It’s a discussion. Is how could this come about … So for me the way I think about it is now, if you will, there’s a bit of a mystery. I see that statistically significant effect in favour of non-price attributes – how can that be? We’ve had a discussion here today.
And I think there’s no disagreement about – that Trivago does not use non-price attributes to decide how to make a decision in regard to the top position offer. And as you think about it, there’s only one conclusion that makes any sense. The only conclusion that makes any sense – as Trivago has also stated – but it’s, indeed, the only conclusion that is consistent as an explanation with the data that I have observed – is that the cost per click – the bid is conveying, implicitly, information that is carrying this idea that the offer might be favourable in other ways. And that is how I believe these non-price attributes are coming into the Algorithm. And I think that’s quite clever.
This is an example of an incentive aligned way to get at information that would not be available to consumers otherwise. The CPC is conveying information in this case about the click to book rate on the site. And it’s information that would not otherwise be visible. And, in this case, through the design of the marketplace, the bids are in an incentive aligned way, it’s aligned with the incentives of the [Online Booking Site]. The bids are conveying this information.
(Emphasis added.)
95 Importantly for present purposes, the above passage makes it clear that Trivago’s Algorithm did not use non-price attributes to determine the composite score (and thus the Top Position Offer). This was common ground between the experts. Professor Parkes nonetheless inferred that the CPC was conveying information about the quality of the underlying offer, and that this was a desirable by-product of the CPC bid mechanism.
96 While Mr Bajanov accepted that the results of Professor Parkes’ analyses were statistically significant, he did not accept that such an inference should be drawn. Mr Bajanov stated in the Joint Report that the experts had no “visibility” as to the way in which Online Booking Sites “estimate or predict their prospective conversion rates in deciding on a CPC bid”. Therefore, in Mr Bajanov’s view, the experts could not make “any definitive statements about the role or importance of these particular non-price attributes”.
97 The primary judge accepted, correctly, that for this reason given by Mr Bajanov, it could not be inferred that the CPC is conveying information about the quality of the underlying offer.
98 The primary judge also concluded that, in any event, when comparing corresponding pairs of Top Position Offers and lowest priced offers, Professor Parkes did not seek to value the non-price attributes. His analysis did not, therefore, provide a basis to compare the relative values of the Top Position Offers and the lowest priced offers. The analysis also did not include offers that were excluded from the listings because they did not meet the minimum CPC threshold.
99 Professor Parkes’ analysis did not, therefore, compare the Top Position Offer with all other offers; it merely compared the Top Position Offer with the cheapest offer that was displayed. For these reasons, the primary judge concluded that the analysis cited above by Professor Parkes did not show that where the Top Position Offer was not the cheapest offer, it nonetheless had some other characteristic which made it more attractive than any other offer for that hotel. As will be seen, Trivago contends that this approach was wrong in principle, and also that the primary judge reversed the onus of proof. This will be addressed below.
5.2 Behavioural evidence – Professor Slonim
100 There was also expert evidence in relation to consumer behaviour. The ACCC sought to rely on a report of Professor Slonim dated 17 May 2019. The whole of the report was subject to objection by Trivago. The primary judge allowed some objections and having done so, commented on the remaining admissible consumer behaviour evidence.
101 Professor Slonim’s report and opinions related only to recreational (as opposed to business) travellers. Further, one of the assumptions Professor Slonim made was that consumers using the Trivago website paid for hotel accommodation at the time of making an online booking. This was not always the case. In some cases, the Top Position Offer was displayed on the Trivago website together with the words “Pay at the hotel”, indicating that the consumer could pay later. Professor Slonim’s evidence was to be understood subject to this qualification.
102 Professor Slonim expressed the following opinions, which the primary judge accepted:
(a) a traditional economics approach, as presented in most basic texts and university courses, assumes that consumers are “maximizers” who gather all available information on all available options and all attributes of every option (e.g., for a hotel room this could include price, amenities, location, services, room size, cancellation policies, applicable taxes, and so forth) to come to a rational (i.e., optimal) decision that maximizes preferences across all attributes. Although this assumption on rational decision-making might explain some macro level market behaviour, there was extensive research with robust evidence over the past 40 years showing that this rational decision-making approach was not descriptive of how most consumers in most contexts make decisions;
(b) in general, consumers were often time constrained, did not have all (or even most) information easily accessible, and could not easily process more than a small amount of information in making choices;
(c) these constraints resulted in consumers using short cuts (known as heuristics) to make choices, and these heuristics were subject to many biases, resulting in deviations from making optimal choices. These biases could be influenced by many seemingly irrelevant factors including, but not limited to: how information was presented, the order in which information was presented, and how choices were framed. The two most important, relevant and well-recognised biases were “present bias” and “loss aversion”;
(d) present bias is a bias in decision-making in which consumers place an inordinate amount of weight on factors that affect them in the present (whether it is a benefit or a cost). Traditional economic analyses have been well-aware for at least a century that consumers might rationally place somewhat more weight in the evaluation of choices on factors that occur sooner rather than later, but present bias indicates that the weight that people place on the present is quite dramatically more important in decisions than the traditional economic approach has considered;
(e) one important consequence of present bias for decision-making in general, and for booking a hotel in particular, was that since the costs of the time spent to find a room were borne at the time of making the decision (i.e., the present), whereas the benefits of finding the room occur in the future when the consumer gets to enjoy the room, consumers will look for ways to expedite the search process by taking shortcuts in the decision process to save time in the present;
(f) loss aversion is a bias in which consumers are assumed to make choices by comparing options with one item serving as the reference (or status quo or default) option and all other choices serving as the alternative options. The loss aversion bias in decision-making explains that in making comparisons to a reference option, an equally sized gain and an equally sized loss were treated quite differently in terms of evaluating each alternative option. This was demonstrated by considering two cases. Case 1: the reference hotel has a price of $180 and an alternative hotel has a price of $200, the alternative hotel option thus has a loss of $20 on price compared to the reference hotel. Case 2: the reference hotel has a price of $220 and an alternative hotel again has a price of $200, the alternative hotel in this case has a gain of $20 on price compared to the reference hotel. In both cases, there is a change in price of $20 between the alternative and the reference hotel;
(g) in Case 2, the relative gain of $20 to the consumer would factor positively on the consumer choice for the alternative hotel, but in Case 1 the $20 loss will weigh much more heavily into adversely affecting the choice for the alternative hotel than the gain would help in Case 2. As a consequence of loss aversion, an option that becomes the reference item would be much more likely to be chosen when there were various trade-offs across attributes, since every alternative would have some attributes that involve a loss compared to the reference item, and these losses weigh more heavily than any gains on other attributes in the evaluation of each alternative option;
(h) another critical aspect in the process of choosing accommodation was the flexibility consumers have when considering the attributes of the accommodation. The attributes of a hotel purchase range from those that are very inflexible, to those that consumers have some ability to compromise on. Inflexible attributes typically include the number of travellers, the general destination (e.g., a city or possibly a neighbourhood within a city) and the date(s) of travel. These inflexible attributes are often decided well before the booking process begins and are ones the traveller is likely unwilling to adjust regardless of the deal offered. The central importance of these attributes was clear given that most Online Booking Sites (e.g., Booking.com, Hotels.com, Expedia.com) ask for these details on their Landing Page;
(i) besides inflexible attributes, there may be several additional attributes of some importance to consumers; consumers might be able to make trade-offs over these additional attributes to some degree assuming they can still satisfy the inflexible attributes. For instance, the attributes that consumers might make trade-offs over include price, room and bed configuration, ratings, hotel amenities, and location within their target destination;
(j) another critical factor affecting online hotel searches for most consumers was that people have a limited ability and motivation to process many choices, more generally referred to as information overload. One approach to dealing with having many options that rational decision-makers might follow, if an initial search produces too many choices, was to use filtering tools (available on most booking sites) to narrow the consideration set. Most websites offer a variety of ways to fine tune the choice set depending on which of these more flexible features are of most importance to the consumer. For instance, consumers can almost always sort on price. Given some evidence suggests that location, price and quality are the three most important attributes to consumers searching for a hotel, to the extent that consumers used filters to refine and narrow their search, it was most likely that consumers would limit their search based on one or more of the three top attributes; and
(k) most people (including those referred to in Professor Slonim’s report as “satisficers”) who faced too many options, rather than proceeding to a more intense search, instead succumb to decision paralysis and accept the first choice that meets their inflexible attribute needs (location, dates, number of people) and is acceptable (i.e., satisfies) their other preferences. For instance, if the price is within some range of acceptability (e.g., under $200 or under $300) and the room has two beds, then they will be satisfied with the choice and proceed to book the room. There are several behavioural reasons this is likely to occur. First, once a consumer finds an initial accommodation that is satisfactory, any rooms that might be found with a further search are unlikely to be evaluated as better due to loss aversion given that there is a high likelihood that even if some attributes are better with the alternative room, other attributes are likely worse, and thus a loss on these attributes will receive greater weight in the evaluation. Secondly, consumers in many cases may not believe that they will be able to find a better deal, either because they do not think that they have the skill, or that it will take too much time to find a potentially better deal (present bias), or that there is not a better deal to be found if they search further. Moreover, in many decision-making situations, consumers often stay with defaults such as the first option they become aware of that is satisfactory (known as a status quo bias) because they lack confidence that another option might actually be a better deal.
103 In relation to the Strike-Through Price, Professor Slonim expressed the following opinions, which the primary judge accepted:
(a) the use of a comparison price that the consumer is not meant to choose is called a “merchant supplied reference price”. This price is used to frame the Top Position Offer price as less expensive by comparison;
(b) more generally, presenting a higher priced alternative to the option the merchant hopes the customer will purchase falls within a broad category of consumer behaviour, psychology and marketing known as the decoy (or attraction) effect when there may be many options to choose from;
(c) the decoy effect influences consumers by making the merchant’s desired item to sell, here the Top Position Offer, look better in comparison to the decoy option;
(d) by positioning the higher priced option in red next to the Top Position Offer, the strike-through reinforces the advantage of the best offer as being a lower price;
(e) another potential effect of the strike-through is that it can help the consumer feel good about clicking through on the Top Position Offer by giving them the belief that they have found a good deal;
(f) given the Strike-Through Price appears to be much higher than the Top Position Offer (often substantially more than $100 higher), the fact that the comparison price is substantially higher will be very likely to lead consumers to think the Top Position Offer is an especially good offer;
(g) once consumers believe the Top Position Offer is an especially good deal compared to the much higher comparison price, they will not believe there is any reason to search further (i.e., they will be satisfied) and will thus proceed to click through to obtain the incorrectly perceived “good deal”; and
(h) since many firms provide merchant supplied reference prices that are usually for the identical item (e.g., furniture), consumers will be used to believing that this much higher comparison price option will be for the same room.
5.3 Economic evidence – Mr Smith
104 Trivago sought to rely on economic evidence from Mr Smith. The primary judge gave reasons for rejecting that evidence in part. Those reasons are not directly challenged on appeal. In dealing with the remaining admissible economic evidence, the primary judge noted that the questions that Mr Smith was instructed to address in his report were as follows:
(a) Are there any economic grounds to conclude that the CPC represents, or is a proxy for, characteristics that are attractive to or beneficial for consumers?
(b) If the answer to the above question is yes, what are those grounds and how (if at all) do they explain the relationship between advertisers’ CPC and the characteristics that are attractive or beneficial to consumers using the Trivago website?
105 Mr Smith dealt with the topic of online advertising and explained that the Trivago website was an example of a platform that operated in a “two-sided” market, in which two distinct types of users obtained value from interacting with the opposite group, on a single platform. Mr Smith stated that both types of user benefitted from the indirect network effects created by the platform that linked them. In this case, the two types of users were consumers searching for a hotel room, and advertisers looking to fill hotel rooms. Mr Smith was not challenged on this description of the Trivago website and the primary judge accepted it.
106 Mr Smith stated that Trivago “provide[d] advertising that [was] free to consumers (consumers [did] not pay to access the Trivago website), but charge[d] advertisers (through the CPC)”. In the course of his oral evidence, the question whether the Online Booking Sites were placing “advertisements” on the Trivago website was raised with Mr Smith. The primary judge had difficulty with the description of the Online Booking Sites as placing “advertisements” on the Trivago website, at least insofar as this description pertained to the perspective of consumers using the website. His Honour noted that this description did not reflect the way in which the Trivago website was presented to consumers in Trivago’s television advertisements. These television advertisements conveyed that Trivago was providing a service for consumers by aggregating prices for hotel accommodation from across the internet, thereby enabling consumers to compare those prices.
107 The television advertisements did not suggest that the Online Booking Sites pay Trivago to have their prices appear on the website. Further, the way in which offers are displayed on the Trivago website did not provide any indication that the Online Booking Sites were placing “advertisements”. The primary judge doubted that many consumers would interpret the offers that appear on the Trivago website as “advertisements”.
108 To this, Mr Smith stated: “Advertising may provide consumers with additional information on available offers and prices”. He also stated: “Separately, advertising costs may act as a signal of the quality of the good being provided.” Mr Smith was cross-examined extensively on these propositions. Mr Smith explained that there were two relevant mechanisms. The first mechanism was just providing information. As Mr Smith put it: “It tells you, Here I am. This was my brand, this is my offer, and this is my price.” The second mechanism puts the goods in the consumer’s mind. As Mr Smith put it, this mechanism “gives you some recognition and some likelihood that you will (a) go and purchase that, and (b) there might be repeat purchases”. His Honour was not satisfied as to this evidence and for the following reason, understandably so. His Honour said (at [173]):
It may be accepted that the display of an Online Booking Site’s offer on the Trivago website performs an information function. However, it is difficult to see how the “advertising costs” associated with an offer (i.e., the CPC) may act as a “signal” to consumers of the quality of the offer in circumstances where consumers using the Trivago website do not know the amount of the CPC. That conclusion is reinforced by the fact that, as Mr Smith accepted, the consumer might not even know there is a CPC.
109 Mr Smith considered whether the CPC “might form a proxy for some characteristics, in particular elements of quality, which are hard to measure objectively, but are still likely to be valuable to consumers”. Mr Smith summarised his conclusions:
I consider that there is substantial differentiation across the offers of different advertisers, even for the same hotel, in regard to the specific features of the room being offered, and various quality elements of the advertisers’ websites. Customers are likely to be aware of, and respond to, these quality differentials. Trivago attempts to control for some aspects of the quality of advertisers’ websites, through a Landing Page score (“LPS”). CPC is likely to provide incremental information over and above LPS, and CPC is likely to be related to, and act as a proxy or signal for, several of the factors that are related to the quality and attractiveness of the advertisers’ websites and offers. The inclusion of CPC, as an additional signal of quality, is likely to allow Trivago’s Algorithm to improve the ranking of higher quality offers, and reduce the ranking of lower quality offers. This would be directly beneficial to consumers.
110 That there was “substantial differentiation across the offers of different advertisers, even for the same hotel” was accepted. Insofar as the above passage contained statements about the Landing Page score, Mr Smith accepted that he had no personal knowledge of how the Landing Page score operated.
111 But, importantly, Mr Smith said that the proposition that the CPC was likely to act as a proxy for several of the factors related to “the quality and attractiveness of the advertisers’ websites and offers” was a matter of economics; “the CPC that is bid for a given advertisement, should be guided by the likely returns from achieving a click on that advertisement”. Mr Smith stated that this might be driven by three factors:
(a) the likely conversion rate – that is, the likelihood that a click through to the Online Booking Site’s website would result in an actual booking;
(b) the margin applicable on conversion to a booking; and
(c) the prospect of an Online Booking Site benefiting from encouraging a consumer to make a repeat visit directly to its website.
112 The primary judge was not persuaded that these three matters, separately or in combination, could possibly support the relevant proposition that the CPC was likely to act as a proxy for the quality and attractiveness of an Online Booking Site’s website and offer.
113 It may be accepted, his Honour said, as a matter of simple economics that the amount that an Online Booking Site was prepared to bid as the CPC was likely to reflect its expected gain. However, there was no reason to presume that an Online Booking Site expecting a higher gain necessarily, or even ordinarily, has a higher quality offer. Mr Smith did not undertake any analysis, using actual data, to see whether or not a higher CPC was associated with a higher quality website and/or offer. Nor did Mr Smith examine Trivago’s Algorithm. Mr Smith did not differentiate between the quality of the Online Booking Site’s website and its offer. In light of these matters, the primary judge gave this aspect of Mr Smith’s report very little weight.
114 Trivago takes exception to this, but the primary judge was entirely entitled to reach this conclusion. His Honour’s reasoning was undoubtedly correct.
6. FINDINGS ON THE REPRESENTATIONS AT TRIAL
115 As to the first relevant sub-period, the ACCC alleged:
(a) the Cheapest Price Representation;
(b) the Strike-Through Representation;
(c) the Top Position Representation; and
(d) the additional conduct allegations.
116 Trivago made admissions in relation to the ACCC’s case based on the Cheapest Price Representation and the Strike-Through Representation in the first relevant sub-period. The primary judge thought it appropriate to deal briefly with these parts of the ACCC’s case, as they provided context for consideration of the other issues, and because Trivago’s admissions did not precisely mirror the ACCC’s case.
6.1.1 Cheapest Price Representation (admitted)
117 It was alleged that Trivago represented in online and television advertising that the Trivago website would quickly and easily identify the cheapest rates available for a hotel room responding to a consumer’s search. This Representation was alleged only in respect of the first relevant sub-period.
118 As noted above, Trivago’s advertising and marketing conveyed the following information to consumers during the first relevant sub-period:
(a) in the television advertisement which aired between 10 January 2016 and 26 September 2017, it was stated that “Trivago makes it easy for you to find the ideal hotel at the best price” and that “Trivago shows you all the different prices for the exact same room, and that’s how you can be sure that you find your ideal hotel for the best price”;
(b) in the television advertisement which aired in the period 29 June 2017 to 1 April 2018, it was stated that “Trivago compares prices from more than 200 websites to make sure you find your ideal hotel for the best price”; and
(c) in the television advertisement, which aired between 19 December 2017 and 1 April 2018, the consumer was urged to “use Trivago to compare these offers and find your ideal hotel for the best price”.
119 His Honour noted (at [198]-[205]) that:
(a) in each case, when the words “best price” were seen in context, it was clear that they referred to the cheapest price (rather than, for example, the best price taking into account the particular features of the various offers). In this context, the words “best price” would be understood by the ordinary consumer to mean the cheapest price;
(b) further, the television advertisements indicated that the Trivago website would quickly and easily help the consumer identify the best price. That was the whole point of the Trivago website: for example, Trivago “does the work for you and instantly compares the prices”;
(c) there was no reason to think that the above two observations did not apply equally to the other television advertisements;
(d) the Google snippet referred both to “the ideal hotel at the best price” and finding the “cheapest price”;
(e) in light of the above, and also having regard to the admissions made by Trivago, it was established that Trivago made the Cheapest Price Representation during the first relevant sub-period. Various changes were made to the Trivago website from 5 December 2017, but these changes did not affect the conclusion that the Cheapest Price Representation was made;
(f) the primary judge said (at [203]):
In making the Cheapest Price Representation during the first relevant sub-period, Trivago engaged in conduct that was misleading or deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. That is because the Trivago website did not quickly and easily identify the cheapest rates available for a hotel room responding to a consumer’s search. The Trivago website does not display offers unless the Online Booking Site’s CPC bid exceeds a minimum threshold set by Trivago. Accordingly, in at least some cases, the cheapest offer for the hotel room did not appear on the Trivago website. This fact alone was sufficient to render the Cheapest Price Representation misleading or deceptive (or likely to mislead or deceive).
(g) further, his Honour said that the expert evidence established that the offer that was given most prominence on the website (that is, the Top Position Offer) was in many cases not the cheapest offer for the hotel room. Based on the data they examined, the computer science experts agreed that higher priced offers were selected as the Top Position Offer over alternative lower priced offers in 66.8% of listings. Conversely, 33.2% of listings had a Top Position Offer that was the cheapest offer. (These figures are based on both the offers that were displayed on the website and the offers that were not displayed because they did not meet the minimum CPC. This is because the dataset provided to the computer science experts included all offers, including those that did not meet the minimum CPC.) The explanation for the fact that in many cases the Top Position Offer was not the cheapest offer related to the role of the CPC in Trivago’s Algorithm. Although the computer science experts differed in their responses to question 3, even on Professor Parkes’ approach, the CPC was a very significant factor in determining the Top Position Offer; it was the second most important factor, with a relative importance of between 33.8% and 44.8%; and
(h) in making the Cheapest Price Representation during the first relevant sub-period, Trivago also engaged in conduct that was liable to mislead the public as to the nature, characteristics and suitability for purpose of the accommodation search service provided by the Trivago website, and thus contravened s 34 of the Australian Consumer Law.
6.1.2 Strike-Through Representation (admitted)
120 The terms of the Strike-Through Representation were that Trivago represented that the Strike-Through Price was a comparison between prices offered for the same room category in the same hotel.
121 The Strike-Through Price appeared in red strike-through text, immediately above the Top Position Offer. It was a similar size to the Top Position Offer.
122 The primary judge reasoned that the implicit representation that was conveyed was that the two offers that were juxtaposed, namely the Strike-Through Price and the Top Position Offer, were comparable offers apart from price; in other words, that the offers were, apart from price, “like-for-like”. This was the natural inference, he said, from the presentation of the Trivago website. It was also supported by Professor Slonim’s evidence, his Honour said. The implicit representation was accentuated by the Percentage Savings box (when this appeared), but the inference arose even when that box did not appear. Further, at least until 29 March 2018, there was no indication on the Trivago website that the offers were not comparable (apart from price). In light of these matters, his Honour considered that, at least until 29 March 2018, Trivago represented that the Strike-Through Price was a comparison between prices offered for the same room category in the same hotel (at [207]). As noted above, on or about 29 March 2018, a hover-over was added in respect of the Strike-Through Price. If a consumer’s mouse cursor hovered over the Strike-Through Price, the following text was displayed:
The strike-through price corresponds to the cheapest offer we received from the most expensive booking site on Trivago for this hotel and your stay dates.
123 His Honour said (at [209]) there was a question whether any statement contained in the hover-over would be sufficient to dispel the impression that the Strike-Through Price and the Top Position Offer were comparable offers (apart from price), because consumers may not place their mouse cursor over the Strike-Through Price and thus may not see the hover-over text.
124 In any event, his Honour considered the wording of this hover-over was confusing, and did not dispel the impression that the Strike-Through Price and the Top Position Offer were comparable offers (apart from price). Importantly, the hover-over did not actually state that the Strike-Through Price may relate to a different room category. In light of these matters, even after 29 March 2018, his Honour concluded that Trivago represented that the Strike-Through Price was a comparison between prices offered for the same room category in the same hotel (at [209]). In making the Strike-Through Representation during the first relevant sub-period, Trivago engaged in conduct that was misleading or deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law.
125 Trivago admitted that the Strike-Through Price did not always relate to the same room category as the Top Position Offer. Trivago also admitted that its conduct in this regard contravened s 18. For the same reasons, his Honour held (at [210]-[211]) that in making the Strike-Through Representation during the first relevant sub-period, Trivago, in connection with the supply or possible supply of accommodation services, made misleading representations with respect to the price of those services, and thus contravened s 29(1)(i) of the Australian Consumer Law. Trivago had also admitted that its conduct in this regard contravened s 29(1)(i).
6.1.3 Top Position Representation (denied)
126 Trivago denies making the Top Position Representation. The alleged representation had two limbs: Trivago represented that the Top Position Offers were either the cheapest available offers (the first limb) or they had some other characteristic which made them more attractive than any other offer for that hotel (the second limb).
127 The primary judge said (at [213]) that in determining whether that representation was conveyed by the Trivago website, it was necessary to consider the online context as well as the formatting (including fonts and colours) used by Trivago on its website. His Honour noted that the Top Position Offer was presented in the far right column, in green, and a relatively large font (compared with the other offers displayed on the Initial Search Results Page). It had white space around it and there was a green “view deal” button below it. The green colour, his Honour said, had positive associations, suggesting that the consumer should go ahead with this offer. The overall impression was that the Top Position Offer was the best offer for the hotel, either in terms of price or some other characteristic. At least until 6 October 2017, there was no contrary indication on the Trivago website. In light of these matters, his Honour considered (at [213]) that, at least until 6 October 2017, Trivago represented that the Top Position Offers were the cheapest available offers for an identified hotel, or had some other characteristic which made them more attractive than any other offer for that hotel. Additional support for this finding was provided by some of the internal Trivago documents which indicated an appreciation of the significance of formatting (including fonts and colours). For example, a Trivago employee wrote in an email that was produced in evidence before the primary judge (and reproduced on appeal) that: “We know as a fact that changing the layout of prices, be it font sizes or colours, has a significant impact on user behaviour and, thus, on conversion.”
128 On about 6 October 2017, the Trivago website began displaying the Our Recommendations information button with a hover-over. If a consumer’s mouse cursor hovered over the Our Recommendations information button, the hover-over would display text stating:
Trivago gives you the opportunity to compare different offers for hotel rooms that are available through Trivago’s website. In some cases, lower prices might be available from other sources. In determining the price to display in the leading position of our search results, we consider a variety of factors, including price, the likelihood that you will find your ideal hotel, your ability to complete a booking after you click on a search result and the level of compensation provided by the booking sites we cover. In order to make more information available on pricing options to our users, additional prices are listed in the “More deals” slide-out.
(Emphasis added.)
129 From about 27 January 2018, the text displayed if a consumer’s mouse cursor hovered over the Our Recommendations information button stated:
The ranking results reflect your search criteria and our assessment of the attractiveness of the offer compared to other offers available on our site. It also reflects the compensation paid by the booking site.
(Emphasis added.)
130 The primary judge (at [216]) again said that there was a question whether any statement contained in the hover-over would be sufficient to dispel the impression created by the presentation of the Top Position Offer on the website because consumers may not place their mouse cursor over the Our Recommendations information button and thus may not see the hover-over text.
131 In any event, his Honour considered the explanation provided by these hover-overs to be “opaque” and insufficient to dispel the impression created by the presentation of the Top Position Offer. The hover-overs referred to the “compensation paid by the booking site”. The hover-overs did not explicitly state that this compensation was being paid to Trivago. Moreover, his Honour observed that the word “compensation” does not readily convey the nature of the CPC mechanism, whereby an Online Booking site was required to pay a fee to Trivago each time a consumer clicks on one of its offers. Finally, the hover-overs did not clearly disclose the significance of the CPC in the selection of the Top Position Offer. Trivago challenges these findings.
132 Apart from the introduction of the hover-over, another change Trivago made to the website during the first relevant sub-period was to the More Deals button. On or about 5 December 2017, the More Deals button ceased to show the number of further offers contained within the More Deals slide-out and instead displayed the price of the cheapest offer contained in the More Deals slide-out. In circumstances where this figure was less than the Top Position Offer, a consumer would be able to determine from the face of the listing that the Top Position Offer was not the cheapest offer for the hotel. On this basis, Trivago submits that it did not make the Top Position Representation after 5 December 2017. The primary judge rejected this submission. First, assuming that the figure in the More Deals button was lower than the Top Position Offer, the figure appearing in the More Deals button was in a much smaller font than the Top Position Offer. It was also in black, rather than green, and at the bottom of a column containing other figures. Having regard to the way in which the figures were presented, his Honour (at [217]) held that “some, perhaps many, consumers” would not have observed the figure in the More Deals button and appreciated that it was less than the Top Position Offer. Secondly, of the consumers who did observe the figure in the More Deals button and did appreciate that it was less than the Top Position Offer, “some, perhaps many”, would have nonetheless understood the Top Position Offer as being either the cheapest offer or having some other characteristic which made it more attractive than any other offer for the hotel (emphasis added). That impression was invited, his Honour said, by the colour, relative size and positioning of the offers. Trivago strongly contends that assessment of the consumer cohort by reference to this measure is erroneous.
133 His Honour concluded (at [218]) that, even after the introduction of the hover-over (in connection with the Our Recommendations information button) and the change to the More Deals button, Trivago represented that the Top Position Offers were the cheapest available offers for an identified hotel, or had some other characteristic which made them more attractive than any other offer for that hotel. This conclusion drew additional support from the click-through figures, which show that there were 4,726,241 sessions where the More Deals slide out was displayed, compared with 20,039,530 sessions where the Top Position Offer for any hotel listing was clicked.
134 His Honour held (at [219]) that in making the Top Position Representation during the first relevant sub-period, Trivago engaged in conduct that was misleading or deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. That was because, contrary to the Top Position Representation, in many cases the Top Position Offer was not the cheapest offer for the hotel, nor did it have some other characteristic that made it more attractive than any other offer for the hotel. As noted above, the Trivago website did not display offers unless the Online Booking Site’s CPC bid exceeded a minimum threshold set by Trivago. Accordingly, in at least some cases, the cheapest offer for the hotel was not actually displayed on the Trivago website.
135 His Honour also reiterated (at [220]) that the expert evidence established that higher priced offers were selected as the Top Position Offer over alternative lower priced offers in 66.8% of listings. Accordingly, in many cases, the Top Position Offer was not the cheapest offer for the hotel.
136 On the second limb of the Top Position Representation, his Honour held that it may be inferred that in at least some cases where the Top Position Offer was not the cheapest offer, it did not have some other characteristic or characteristics that made it more attractive than any other offer for the hotel. His Honour found that Trivago’s Algorithm did not use non-price attributes of the offers to determine the composite score (and thus the Top Position Offer). A very significant factor in the selection of the Top Position Offer was the Online Booking Site’s CPC bid, rather than the quality of the hotel accommodation offer itself. Insofar as Professor Parkes analysed certain cases where the Top Position Offer was at least 1% more than the cheapest offer and drew an inference that the CPC was conveying information about the quality of the underlying offer, the primary judge was not satisfied that such an inference should be drawn. Further, Professor Parkes did not seek to value the non-price attributes, and did not consider offers that were excluded because they did not meet the minimum CPC. His analysis did not, therefore, provide a basis to compare the relative value of the Top Position Offer and the lowest priced offer. Insofar as Mr Smith expressed the view, based on economic theory, that the CPC is likely to act as a proxy for factors that are related to the quality and attractiveness of the offer, the primary judge gave that opinion very little weight (at [221]). Trivago submitted that the correlation between the priority modifier and the Landing Page score allowed the quality of the Online Booking Site’s website to be taken into account. However, the evidence did not establish how the Landing Page score in fact operated in practice. For this reason, his Honour was not prepared to infer that it operated as a quality measure in this way in the absence of evidence from a person with actual knowledge about its operation. Further and in any event, this point at best related to the quality of the website rather than to the quality of the hotel accommodation offer. The second limb of the Top Position Representation was concerned with the hotel accommodation offer rather than with the quality of the Online Booking Site’s website.
137 The primary judge concluded (at [222]) that for the same reasons, in making the Top Position Representation during the first relevant sub-period, Trivago, in connection with the supply or possible supply of accommodation services, made misleading representations with respect to the price of those services (and thus contravened s 29(1)(i) of the Australian Consumer Law).
6.1.4 Additional conduct allegations
138 The ACCC alleged that, by making the Cheapest Price Representation, and by making Top Position Offers together with the Top Position Representation and the Strike-Through Representation, during the first relevant sub-period, Trivago led consumers to believe that the Trivago website provided an impartial, objective and transparent price comparison which would enable them to quickly and easily identify the cheapest available offer for a particular (or the exact same) room at a particular hotel.
139 Having concluded that Trivago made the Cheapest Price Representation, the Strike-Through Representation and the Top Position Representation during the first relevant sub-period and considering this conduct as a whole, his Honour was satisfied (at [224]) that Trivago did lead consumers to believe that the Trivago website provided an impartial, objective and transparent price comparison which would enable them to quickly and easily identify the cheapest available offer for a particular (or the exact same) room at a particular hotel. For example, in the television advertisement that aired from 10 January 2016 to 26 September 2017, the presenter referred to the fact that there are so many different prices out there (i.e. online) “for the exact same room”. Trivago “does the work for you and instantly compares the prices of over 600,000 hotels from over 200 different websites”. It was said that, instead of “searching for hours”, Trivago “makes it easy for you” to find the ideal hotel for the best price. The presenter says “just go to Trivago, type in where you want to go, and with two clicks select your check-in and check-out dates and search; It’s that simple”. It was said that Trivago shows you all the different prices “for the exact same room”. The overall impression was that Trivago was impartial, objective and transparent. Although not as lengthy and detailed, the other television advertisements were to similar effect. The Strike-Through Representation and the Top Position Representation were consistent with the messages conveyed in the television advertisements. While changes were made to the Trivago website during the course of the first relevant sub-period, these were not sufficient to dispel the impression created by the television advertisements, according to the primary judge.
140 The primary judge concluded (at [225]) that by engaging in the conduct described in the preceding paragraph during the first relevant sub-period, Trivago engaged in conduct that was misleading and deceptive or likely to mislead or deceive (and thus contravened s 18 of the Australian Consumer Law), and conduct that was liable to mislead the public as to the nature, characteristics and suitability for purpose of the accommodation search service provided by the Trivago website (and thus contravened s 34 of the Australian Consumer Law).
141 His Honour said that contrary to the impression created by the relevant conduct, the Trivago website did not provide an impartial, objective and transparent price comparison service. In fact, the selection of the Top Position Offer was heavily influenced by the CPC, which was not made clear. Many consumers would not have appreciated that Trivago was being paid by the Online Booking Sites, let alone that the amount it was being paid was a very significant factor in the selection of the Top Position Offer. Further, the Trivago website did not quickly and easily identify the cheapest rates available for a hotel room responding to a consumer’s search. Importantly, in at least some cases, the cheapest offer for a hotel room was filtered out and therefore did not appear on the Trivago website.
6.2 The second relevant sub-period
142 In respect of the second relevant sub-period (29 April 2018 to 20 November 2018), the ACCC alleged the following representations and conduct:
(a) the Strike-Through Representation;
(b) the Top Position Representation; and
(c) the additional conduct allegations.
6.2.1 Strike-Through Representation (denied)
143 The Strike-Through Representation alleged that Trivago represented that the Strike-Through Price was a comparison between the prices offered for the same room category in the same hotel. The format and appearance of the Strike-Through Price was essentially the same during the second relevant sub-period as during the first relevant sub-period. Accordingly, at least until August 2018, the reasoning was equally applicable.
144 In August 2018, the text that was displayed when a consumer’s mouse cursor hovered over the Strike-Through Price was amended to state:
This is the cheapest deal from the most expensive booking site with offers for this hotel on your stay dates.
145 The primary judge said (at [230]) that, again, there was a question whether any statement contained in the hover-over would be sufficient to dispel the impression that the Strike-Through Price and the Top Position Offer were “like-for-like” offers (apart from price). The wording of the hover-over, he said, was also confusing, and failed to dispel this impression. In light of these matters, even after the change in the wording of the hover-over, his Honour held that Trivago represented that the Strike-Through Price was a comparison between prices offered for the same room category in the same hotel.
146 In making the Strike-Through Representation during the second relevant sub-period his Honour held (at [231]) that Trivago engaged in conduct that was misleading or deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. Trivago admitted that the Strike-Through Price did not always relate to the same room category as the Top Position Offer. Trivago has admitted that its conduct in this regard contravened s 18.
147 For the same reasons, in making the Strike-Through Representation during the second relevant sub-period, Trivago, in connection with the supply or possible supply of accommodation services, made misleading representations with respect to the price of those services (and thus contravened s 29(1)(i) of the Australian Consumer Law). The primary judge noted that Trivago had also admitted that its conduct in this regard contravened s 29(1)(i).
6.2.2 Top Position Representation (denied)
148 The primary judge noted (at [234]) that the way in which listings appeared on the Trivago website during the second relevant sub-period was similar to the first relevant sub-period, save that the one difference was the change to the wording on the Landing Page. From 12 April 2018 (i.e. just before the start of the second relevant sub-period), the words on the Landing Page were amended to read: “Find your ideal hotel and compare prices from different websites”. In addition, from 3 July 2018, a link called “Learn how Trivago works” was introduced. This provided information about Trivago’s Algorithm and the ranking of offers on the Initial Search Results Page. In about October 2018, a link to the “Learn how Trivago works” page appeared in the hover-over for the Our Recommendations information button.
149 The primary judge held (at [235]) that, for the same reasons, the overall impression was that the Top Position Offer was the best offer for the hotel (either in terms of price or some other characteristic). The discussion of the hover-overs was equally applicable in respect of the second relevant sub-period. The discussion of the More Deals button was also equally applicable in respect of the second relevant sub-period.
150 Although the wording of the Landing Page was different during the second relevant sub-period, his Honour did not consider this to affect whether or not the Top Position Representation was conveyed. “The new wording, which refers to comparing prices from different websites, is neutral as to whether or not the representation was made” (at [237]).
151 The primary judge found (at [238]) that, at least until 3 July 2018, Trivago represented that the Top Position Offer was the cheapest available offer for a given hotel, or had some other characteristic which made it more attractive than any other offer for that hotel. The link that was introduced on about 3 July 2018 provided information about Trivago’s Algorithm and the ranking of offers on the Initial Search Results Page. His Honour also considered whether a consumer would be likely to click on the link. It was not clear from the statement of agreed facts where the link appeared on the Trivago website. In light of these matters, even after the introduction of this link, the impression created by the presentation of the Top Position Offer was the same as discussed above.
152 In relation to the link to the “Learn How Trivago Works” page that was included in the Our Recommendations information button from about October 2018, there was a question whether a consumer would be likely to click on the link, since it was within a hover-over. In any event, the explanation provided in the linked page (set out at [46(b) above), which was entitled “How does Trivago determine the ‘our recommendations’ sort”, did not explain in clear terms, the primary judge said, the significance of the CPC in the selection of the Top Position Offer. Rather, its role was downplayed by referring to it last after a number of other factors. The second paragraph of the page stated that the consumer may find offers in the More Deals slide-out with a lower price than the Top Position Offer. However, it did not state that Trivago filters out offers that do not meet its minimum CPC threshold and that these offers, which may be cheaper, did not appear at all. The statement that, “while price is an important factor when selecting the Top Position Offer, we believe other factors, such as those mentioned above, make offers attractive and relevant to you, and contribute to high levels of satisfaction by our users” tended to suggest, his Honour said (at [240]), that, if the Top Position Offer was not the cheapest offer, this was because the offer has non-price characteristics that made it more attractive than the cheapest offer. His Honour said that this was consistent with, and supported, the proposition that Trivago made the Top Position Representation.
153 His Honour concluded (at [241]) that even after the introduction of the links on 3 July 2018 and in October 2018, Trivago still represented that the Top Position Offer was the cheapest available offer for a given hotel, or had some other characteristic which made it more attractive than any other offer for that hotel.
154 In making the Top Position Representation during the second relevant sub-period, Trivago was found to have engaged in conduct that was misleading and deceptive or likely to mislead or deceive. For the same reasons, in making the Top Position Representation during the second relevant sub-period, Trivago, in connection with the supply or possible supply of accommodation services, made misleading representations with respect to the price of those services, and thus contravened s 29(1)(i) of the Australian Consumer Law.
6.2.3 Additional conduct allegations (denied)
155 For the same reasons as in the first relevant sub-period, the primary judge concluded (at [245]) that Trivago made the Strike-Through Representation and the Top Position Representation during the second relevant sub-period. Trivago’s television advertisements continued until 2 July 2018 and conveyed the information and statements to consumers discussed above (at [26]-[31]).
156 In relation to the period after 2 July 2018, there was no evidence that Trivago conducted television advertising. Insofar as the ACCC relied on other forms of advertising:
(a) at [6] of its Amended Concise Statement, the ACCC refers to the snippet that appeared in Google search results between 1 December 2016 and a date that is not known to the ACCC but was not before 14 January 2019 (see [32] above). It was alleged that the snippet stated: “Compare over 250 booking sites and find the ideal hotel at the best price!”; and “Compare hotels, find the cheapest price and guarantee the best deal on accommodation …”;
(b) in [8A] of the Amended Concise Statement, the ACCC stated:
Since on or around 15 April 2018, the Trivago website has stated: “Find your ideal hotel and compare prices from different websites”. Trivago has also caused the following statements to be made in the snippet which appeared beneath search results displayed when Australian consumers conducted an online search for “Trivago”; “Compare 200+ booking sites globally, find the ideal hotel at great price! Find the perfect hotel. Free and Easy to Use. Fast and Simple. 1M+ Hotels Worldwide. High-quality hotel photos. Compare & Save. No Ads or Pop-ups” and “Compare 200+ booking sites globally, find the ideal hotel at great price! Many prices in one site. Save Time & Money. Enjoy Your Stay. Impartial Comparison. Compare Booking Sites. 1M+ Hotels Worldwide. Act Fast for Great Deals.”
(c) insofar as the ACCC relied on snippets, the statement of agreed facts contained an agreed fact relating to a snippet, but it concerned a period before 2 July 2018. Trivago made admissions regarding the snippet, but these also related to a period before 2 July 2018. Trivago admitted that on or about 14 January 2019, it came to its attention that the words “best price” appeared in a subset of Google unpaid Trivago-related Google search results in Australia. However, the admission only went so far as to say that the snippet contained the words “best price”. It did not set out the rest of the wording of the snippet (see [50] above).
157 In respect of the period after 2 July 2018, his Honour was not satisfied (at [247]) that, by making Top Position Offers on the Trivago website, making the Top Position Representation and the Strike-Through Representation, and stating “Find your ideal hotel and compare prices from different websites” on the Trivago website, Trivago led consumers to believe that the Trivago website provided an impartial, objective and transparent price comparison which would enable them to quickly and easily identify the cheapest available offer for a particular (or the exact same) room at a particular hotel. In other words, the additional conduct allegations were not made out in the second relevant sub-period after 2 July 2018.
158 However, by engaging in the conduct up to 2 July 2018 in the second relevant sub-period, the primary judge did find Trivago engaged in conduct that was misleading or deceptive or likely to mislead or deceive (and thus contravened s 18 of the Australian Consumer Law) and conduct that was liable to mislead the public as to the nature, characteristics and suitability for purpose of the accommodation search service provided by the Trivago website (and thus contravened s 34 of the Australian Consumer Law). His Honour considered that the reasoning in relation to the additional conduct allegations in the first relevant sub-period was also applicable to this conduct in the second relevant sub-period up to 2 July 2018 (at [138]-[141]) above).
6.3 The third relevant sub-period
159 In respect of the third relevant sub-period (20 November 2018 to 13 February 2019), the ACCC alleged the following representations and conduct:
(a) the Strike-Through Representation (in respect of the map portion of the website);
(b) the Red Price Representation;
(c) the Top Position Representation; and
(d) the additional conduct allegations.
6.3.1 Strike-Through Representation (denied)
160 The asserted Strike-Through Representation was that Trivago represented that the Strike-Through Price was a comparison between the prices offered for the same room category in the same hotel.
161 The statement of agreed facts established that the Strike-Through Price continued to appear in the map portion of the Trivago website during the third relevant sub-period.
162 The primary judge (at [252]) inferred that the Strike-Through Price appeared immediately above or, at least, close to the Top Position Offer. The implicit representation conveyed by this positioning was that the two offers were comparable apart from price; in other words, that the offers were, apart from price, “like-for-like”. In light of this, his Honour considered that Trivago represented that the Strike-Through Price was a comparison between prices offered for the same room category in the same hotel.
163 The primary judge considered (at [253]) for the same reasons given in relation to the previous sub-periods, that in making the Strike-Through Representation in the map portion of the Trivago website during the third relevant sub-period, Trivago engaged in conduct that was misleading or deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. Trivago admitted that “the room rate [i.e. the Strike-Through Price] did not always relate to the same room category as the Top Position Offer”. While that admission was made in the context of the period up to 20 November 2018 (the date when Trivago ceased using the Strike-Through Price on the Initial Search Results Page), it was apparent that the proposition also applied to the Strike-Through Price in the map portion of the website during the third relevant sub-period. This was inferred from the Red Price hover-over, which included the statement that the Red Price “may relate to a room type or amenities package that differs from other offers on our site”.
164 For the same reasons, the primary judge held (at [254]) that in making the Strike-Through Representation in the map portion during the third relevant sub-period, Trivago, in connection with the supply or possible supply of accommodation services, made misleading representations with respect to the price of those services, and thus contravened s 29(1)(i) of the Australian Consumer Law.
6.3.2 Red Price Representation (denied)
165 This asserted Representation was that Trivago represented that the Red Price was a comparison between prices offered for the same room category in the same hotel, in effect the same representation alleged with respect to the Strike-Through Price in the previous sub-periods (and the map portion in the third relevant sub-period).
166 The Red Price appeared in red text in the same column as the Top Position Offer. No other offers were set out in that column of the listing for the hotel. Although the Red Price was a smaller size than the Top Position Offer, the presentation of both offers in the same column of the listing conveyed that the two offers were comparable offers apart from price; in other words, that the offers were, apart from price, “like-for-like”. This, the primary judge said (at [256]), was the natural inference from the way the two offers were presented. It was also supported by Professor Slonim’s evidence, albeit generally expressed in terms of the Strike-Through Price.
167 Throughout the third relevant sub-period, if a consumer’s mouse cursor hovered over the Red Price, the following text was displayed:
For your stay dates, this is the cheapest deal from the most expensive booking site with offers for this hotel. It may relate to a room type or amenities package that differs from other offers on our site.
168 Again, his Honour said (at [258]) that there was a question whether any statement contained in the hover-over would be sufficient to dispel the impression that the Red Price and the Top Position Offer were comparable offers (apart from price). Further, the wording of the hover-over was confusing, and did not dispel the impression that would otherwise be given that the Red Price and the Top Position Offer were comparable offers (apart from price).
169 In light of the above, and notwithstanding the hover-over, the primary judge concluded (at [259]) that Trivago represented that the Red Price was a comparison between prices offered for the same room category in the same hotel. In making the Red Price Representation during the third relevant sub-period, Trivago engaged in conduct that was misleading and deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. His Honour inferred from the second sentence of the hover-over that the Red Price did not always relate to the same room category as the Top Position Offer. There was no evidence from any officer or employee of Trivago to the contrary.
170 Therefore, for the same reasons, his Honour found (at [261]) that in making the Red Price Representation during the third relevant sub-period, Trivago, in connection with the supply or possible supply of accommodation services, made misleading representations with respect to the price of those services, and thus contravened s 29(1)(i) of the Australian Consumer Law.
6.3.3 Top Position Representation (denied)
171 Apart from the change from the Strike-Through Price to the Red Price, the appearance of the Trivago website was substantially the same during the third relevant sub-period as it had been during the last part of the second relevant sub-period (that is, from about October 2018). The primary judge therefore reached the same conclusions (at [262]-[263]) as set out in that section with respect to the Top Position Representation in the third relevant sub-period.
6.3.4 Additional conduct allegations
172 The same approach to, and findings on, this Representation by the primary judge (at [264]) continued in this period.
6.4 The fourth relevant sub-period
173 In respect of the fourth relevant sub-period (13 February 2019 to 13 September 2019), the ACCC alleged the following representations and conduct:
(a) the Red Price Representation;
(b) the Top Position Representation; and
(c) the additional conduct allegations.
6.4.1 Red Price Representation (denied)
174 The main difference between the third and fourth relevant sub-periods was the removal of the Strike-Through Price from the map portion of the website. This did not affect the reasoning in relation to the Red Price Representation.
175 On about 17 April 2019, during the course of the fourth relevant sub-period, the words “How payments to us affect ranking” were added next to the Our Recommendations information button. The primary judge did not consider the addition of these words to affect the analysis regarding the Red Price Representation (at [270]).
6.4.2 Top Position Representation (denied)
176 Save for the introduction of the words “How payments to us affect ranking” next to the Our Recommendations information button, the appearance of the Trivago website was substantially the same during the fourth relevant sub-period as it had been during the third relevant sub-period.
177 The primary judge considered (at [276]) that while the words added next to the Our Recommendations information button may draw a consumer’s attention to the button, and therefore the hover-over associated with that button, there were deficiencies with the wording of the hover-over and the link included in that hover-over. Accordingly, notwithstanding the addition of those words, the reasoning in relation to the third relevant sub-period, which in turn adopted the reasoning in relation to the period from October 2018 in the second relevant sub-period, was equally applicable. His Honour reached the same conclusions in relation to the fourth relevant sub-period.
6.4.3 Additional conduct allegations (denied)
178 The primary judge considered (at [277]) that the reasoning in respect of this representation in the period after 2 July 2018 in the second relevant sub-period was equally applicable and the same conclusion was reached.
179 There are 17 grounds of appeal, including two new grounds added at the hearing of the appeal. However, they were argued in the following sequence, which these reasons will adopt.
180 Trivago challenges those parts of the judgment and declarations in which the primary judge found and declared that:
(a) during the whole of the Relevant Period, Trivago impliedly made the alleged Top Position Representation, being that the Top Position Offers were the cheapest available offers for an identified hotel or had some other characteristic which made them more attractive than any other offer for that hotel: (at [213], [217], [218], [238], [241], [263] and [276]). (appeal grounds 1 to 5);
(b) the Top Position Representation, if made, was misleading or deceptive because the Top Position Offers (where they were not the cheapest) did not have some other characteristic that made them more attractive than any other offer for the hotel: (at [221]) (appeal grounds 6 to 11);
(c) from 20 November 2018 to 13 September 2019, Trivago impliedly made the alleged “Red Price Representation”, being that the Red Price was a comparison between prices offered for the same room category in the hotel: (at [259] and [271]) . (appeal grounds 12 and 13);
(d) from 1 December 2016 to 2 July 2018, Trivago led consumers to believe that the Trivago website provided an impartial, objective and transparent price comparison which would enable them to quickly and easily identify the cheapest available offer for a particular room at a particular hotel when that was not the case: (at [224] and [225]). (appeal grounds 14 and 15); and
(e) the primary judge drew two important inferences concerning the Red Price Representation which his Honour was not entitled to (additional grounds raised at the hearing).
181 Those grounds are now considered.
8. APPEAL GROUNDS CONSIDERATION
8.1 Grounds 1-5 – Was there a Top Position Representation?
182 Grounds 1-5 challenge the making of the Top Position Representation at all, whereas grounds 6-11 challenge whether, if made, that it was misleading. The ACCC alleges that the Top Position Representation arose from the display of the Initial Search Results Page as to which the primary judge ruled as follows (at [213]):
In determining whether that representation was conveyed by Trivago, by the Trivago website, it is necessary to consider the online context as well as the formatting (including fonts and colours) used by Trivago on the website. The Top Position Offer was presented in the far right column, in green, and a relatively large font (compared with the other offers displayed on the Initial Search Results Page). It had white space around it, and there was a green “View Deal” button below it. The green colour has positive associations, suggesting that the consumer should go ahead with this offer. The overall impression was that the Top Position Offer was the best offer for the hotel, either in terms of price or some other characteristic. At least until 6 October 2017, there was no contrary indication on the Trivago website. In light of these matters, I consider that, at least until 6 October 2017, Trivago represented that the Top Position Offers were the cheapest available offers for an identified hotel, or had some other characteristic which made them more attractive than any other offer for that hotel. Additional support for this finding is provided by some of the internal Trivago documents included in the Court Book, which indicate (as one would expect) an appreciation of the significance of formatting (including fonts and colours) (see, in particular, tabs 77, 85 and 162). For example, in the email at tab 77, an employee of Trivago wrote: “We know as a fact that changing the layout of prices, be it font sizes or colours, has a significant impact on user behaviour and, thus, on conversion.”
183 The primary judge’s approach, Trivago says, fails to apply the correct legal test. Instead, isolated and undue emphasis was placed on certain aspects of the display of the Top Position Offers while the context in which they appeared was ignored, including the available sorting mechanisms, and the clear and prominent display of equal and cheaper offers in the Second, Third or Fourth Position Offers and the More Deals button, and the other information readily available from the webpage.
184 Of central emphasis in Trivago’s appeal was the question of whether the correct test was applied. Trivago contends that the primary judge erred at the outset of his analysis by failing to identify and apply the knowledge base to be attributed to the hypothetical, ordinary and reasonable member of the public looking to book accommodation online. Trivago argues that in this case, the ordinary and reasonable consumer can be expected to have knowledge of the following matters relevant to the alleged representations:
(a) there are various websites which provide online hotel and accommodation booking services, including both aggregator sites such as Booking.com, Hotels.com, Expedia.com, Agoda.com, Wotif.com and the Online Booking Sites themselves, that directly offer hotels or accommodation;
(b) there are various websites, such as Trivago’s website, which aggregate offers from Online Booking Sites. Such sites do not provide booking services;
(c) in order to show accommodation, booking and aggregator sites require consumers to input search parameters which typically include location, number of travellers and dates of travel. These booking and aggregator sites then produce a set of search results showing offers which meet the search parameters;
(d) customers who visit these sites may do so in order to make a booking or to investigate available options, in terms of hotel, hotel attributes, and comparative prices. They would know from past experience that the prices offered by booking and aggregator sites may vary, as may applicable terms and conditions, such as payment date, cancellation policies, free WiFi, etc. They may be interested in accessing multiple sites to investigate and compare details of this kind;
(e) booking and aggregator sites have filtering tools which enable consumers to narrow or order the Initial Search Results Page, such as by reference to “price”, “ratings” or “recommended”;
(f) by clicking on the offers on the Initial Search Results Page, consumers can obtain further details about the offers from other booking sites, including photographs of the hotel and accommodation and non-price features, such as “free breakfast”, “WiFi”, whether the offer is refundable, and whether consumers are required to pay at the time of booking or on check-in. They will also note that they can click through to the actual hotel site, which is likely to provide photographs and information concerning the room, its amenities and the terms and conditions offered by the hotel (for example, cancellation policies, WiFi, the inclusion of breakfast etc); and
(g) consumers can also click on, or hover-over particular buttons or entries on the Trivago Initial Search Results Page to obtain further information. They are able to study the information outlined above while going backwards and forwards between other websites and links as they need.
185 Although the primary judge noted (at [213]) that there were alternative elements to the Top Position Representation, namely, if not the cheapest available offer for that hotel then alternatively, “some other characteristic which made it more attractive than any other available offer for that hotel”, this finding was incorrect, Trivago says. This was because, contrary to the primary judge’s conclusions, the notion that the Top Position Offer was the cheapest offer was defeated by the default sort (for example, our recommendations) and by the display of equal and cheaper offers next to the Top Position Offer. Separately from the More Deals button, equal or cheaper offers to the Top Position Offers were clearly and prominently displayed, Trivago says, in the Second, Third or Fourth Position Offers in the Relevant Period.
186 As to the sorting options on the Initial Search Results Page, Trivago says the primary judge took no account of this. At the hearing, it was put to Trivago that the sorting mechanisms only allowed a consumer to adjust the order of the various listings for different hotels produced by their search. The sorting mechanisms appeared to have no bearing on how different offers for the same hotel listing were displayed. Trivago appeared to accept this proposition with the qualification that the existence of the sort function indicates that Trivago is simply putting forward its recommendations to consumers, not some kind of absolute objective circumstance.
187 As to the position of the Top Position Offer, Trivago argues that the primary judge’s analysis overlooks this matter, and it could not be sensibly argued that the ordinary and reasonable consumer would not notice the equal or cheaper prices displayed in the Second, Third and/or Fourth Position Offers. Trivago says those facts alone defeat the ACCC case to the effect that the Top Position Representation was conveyed by the prominent display of all Top Position Offers on the Trivago website during the Relevant Period.
188 Additionally, for most of the Relevant Period, (from 5 December 2017 to 13 September 2019), equal or cheaper offers to the Top Position Offer were displayed on the More Deals button. The More Deals button clearly and prominently displayed equal or cheaper offers frequently. The fact that the More Deals button often provided for lower prices than the Top Position Offer was clearly stated in the text accessed through the Our Recommendations information button.
189 Trivago says the primary judge wrongly discounted this clear disclosure. It argues that the analysis adopted by his Honour does not withstand scrutiny for several reasons:
(a) first, the primary judge applied the wrong legal test. Rather than considering whether the alleged representation would be conveyed to the ordinary and reasonable member of the target audience, the primary judge approached the matter from the perspective of “some, perhaps many” consumers (at [217]);
(b) secondly, the clear and prominent display of equal and cheaper prices in the More Deals button would not be overlooked by the ordinary and reasonable consumer merely because prices were in a different colour and column to the display of the Top Position Offers. While it may be accepted that the colour green has positive associations, as the primary judge concluded, it does not follow that the ordinary and reasonable consumer would not observe, or would disregard, the adjacent prices or the prices in the More Deals button. Trivago says this is not a fine print case; and
(c) thirdly, Trivago argues the primary judge’s proposition that consumers who observe the display of equal and cheaper prices in the More Deals button would nonetheless have understood that the Top Position Offer was “either the cheapest offer” or had “some characteristic which made it more attractive than any other offer for the hotel” is illogical. It could not be the former, nor could it only be the latter, Trivago says. Once the ordinary and reasonable consumer observed equal and cheaper prices in the adjacent Second, Third or Fourth Position Offers or the More Deals button, he or she could not have implied, and could not have understood, that the Top Position Offer was the cheapest.
190 Trivago’s argument that the primary judge applied the wrong test appears to ignore what his Honour said (at [192]) regarding the relevant class of consumers which is repeated for convenience:
Relevant class of consumers
The relevant class of consumers for present purposes is those members of the public looking to book accommodation online. It may be expected that ordinary and reasonable members of this class of consumers have some familiarity with using the internet and making bookings (whether for accommodation or other products or services) online. I would not assume that such consumers have used the Trivago website before or that they have booked accommodation online before. I note that the relevant class is very large, as indicated by the figures set out in [88] above. While those figures do not state the number of consumers who accessed the Trivago website, they indicate that there were a large number of sessions on the website. For example, in the period 1 December 2016 to 3 January 2018, there were 20,039,530 sessions on the Trivago website where the Top Position Offer for any hotel listing was clicked. The evidence does not establish how long consumers are likely to spend on the Trivago website. The First Parkes Report refers at [131] to a study called “TripAdvisor Insights” which found that the average hotel purchaser spends 191.4 minutes researching online and visits 34 sites before making a purchase. Professor Parkes did not himself carry out a study about this matter and there is a question as to the weight that can be ascribed to it. In any event, I consider it unlikely that many consumers would spend this period of time (or anything approaching it) on the Trivago website. The whole point of the Trivago website, as explained in Trivago’s advertising (see, for example, the television advertisement at [3] above) was to save the consumer time by not needing to search many different websites. Further, Professor Slonim’s evidence regarding ‘satisficers’ suggests that many consumers are likely to interact only briefly with the website.
(Emphasis added.)
191 True it is that the primary judge subsequently (at [217] referred to “some or possibly many” users, but this was clearly a reference to the relevant class of consumers within those identified in the passage above. This reference did no more than eliminate a de minimis quantity. Although his Honour has incorrectly and unnecessarily added a further gloss to the analysis, this has no effect at all on the substance of his Honour’s reasoning. These words must be read in the context of the specific assessment set out at [192].
192 Since the primary judge delivered his decision, the Full Court has (recently) and, with respect correctly, clearly articulated the appropriate test in Australian Competition and Consumer Commission v TPG Internet Pty Ltd [2020] FCAFC 130 per Wigney, O’Bryan and Jackson JJ, where their Honours said (at [23]):
The primary judge referred to a further or alternative test for whether conduct that is directed to the public generally or a section of the public is misleading: whether a significant number of persons to whom the conduct is directed would be led into error (PJ at [41] and [42]). On the appeal, TPG relied on the test, submitting that it is necessary for the ACCC to show that a “not insignificant number” of reasonable persons within the relevant class of persons (prospective purchasers of TPG’s services) have been misled or deceived or are likely to be misled or deceived. No substantive argument was directed to the correctness of that test by the ACCC and our decision in this appeal does not turn upon it. Nevertheless, we consider it appropriate to record our view that the test is, at best, superfluous to the principles stated by the High Court in Puxu, Campomar and Google Inc and, at worst, an erroneous gloss on the statutory provision. We make the following brief observations:
(a) The origin of the “significant number” test can be traced back to decisions of Franki J in Weitmann v Katies Ltd (1977) 29 FLR 336 (at 343) and Wilcox J (as a member of a Full Court) in 10th Cantanae Pty Ltd v Shoshana Pty Ltd (1987) 79 ALR 299 (at 302). In both cases, the test appears to have been formulated by adoption of principles applied in the law of passing off. However, it has long been settled that the proper construction of s 18 is not controlled by analogous common law or statutory remedies: see, for example, World Series Cricket v Parish (1977) 16 ALR 181 at pp 198-199 per Brennan J (as a judge of the Federal Court), referred to with approval in Puxu at p 198 per Gibbs CJ and at 204-205 per Mason J, and Campomar at [97].
(b) The test has never been embraced by the High Court. As early as 1981 in Puxu, the High Court formulated the relevant question as the effect of the impugned conduct on reasonable or ordinary members of the class of persons to whom the conduct was directed: see at 199 per Gibbs CJ and 210 per Mason J. In the High Court cases that have followed Puxu, particularly Campomar and Google, the test has always been stated in substantially the same terms.
(c) The correctness of the “significant number” test was doubted by Finkelstein J at first instance in Australian Securities and Investments Commission v National Exchange Pty Ltd [2003] FCA 955; 202 ALR 24 at [11] and, a short time later in .au Domain Administration v Domain Names Australia Pty Ltd [2004] FCA 424; 207 ALR 521 at [22]-[26], Finkelstein J concluded that the test had been overtaken by the test stated by the High Court in Campomar. In the appeal from the National Exchange decision, the Full Court expressed the view that the “significant number” test was merely an alternative way of expressing the test stated by the High Court in Campomar: National Exchange Pty Ltd v Australian Securities and Investment Commission [2004] FCAFC 90; 49 ACSR 369 (National Exchange) at [23] per Dowsett J and at [70] per Jacobson and Bennett JJ.
(d) Despite the view expressed by the Full Court in National Exchange, the “significant number” test has been referred to and applied in a number of cases in the Federal Court, often in the context of “passing off” type cases. In particular, the test was applied by Tamberlin and Siopis JJ in Hansen Beverage Company v Bickfords (Australia) Pty Ltd (2008) 171 FCR 579 at [45]-[48], by Greenwood J in Bodum v DKSH Australia Pty Ltd [2011] FCAFC 98; 280 ALR 639 at [209] (with whom Tracey J agreed), and by Greenwood, Logan and Yates JJ in Global One Mobile Entertainment Pty Ltd v ACCC [2012] FCAFC 134 at [108], [111]. But these cases did not resolve the question whether it was a different and additional test to the principles stated by the High Court.
(e) More recently, in Flexopack SA Plastics Industry v Flexopack Australia Pty Ltd [2016] FCA 235; 118 IPR 239, Beach J made passing reference to the test, stating that he was inclined to the view that if, applying the Campomar test, reasonable members of the class would be likely to be misled, then such a finding carries with it that a significant proportion of the class would be likely to be misled (at [270]). But later, in Australian Competition and Consumer Commission v Get Qualified Australia Pty Ltd (in liquidation) (No 2) [2017] FCA 709 at [42] and in Australian Securities and Investments Commission v Westpac Banking Corporation (No 2)(2018) 266 FCR 147 at [2279], his Honour said he had become inclined to the view that a finding that reasonable members of the class would be likely to be misled does not necessarily carry with it that a significant proportion of the class would be likely to be misled, so that a finding of a "not insignificant number" of members of the class being likely to be misled is an additional requirement that needs to be satisfied.
(f) Respectfully, we disagree. Whether or not speaking of a reasonable member of a class implies as a matter of strict logical necessity that one is speaking of a significant proportion of that class (cf. Dowsett J in National Exchange at [23]), nothing in the language of the statute requires the court to determine the size of any such proportion. We would emphasise the point that Dowsett J made in National Exchange at [24] that the test of reasonableness involves the recognition of the boundaries within which reasonable responses will fall. In our view it does not require any attempt to quantify, even approximately, the hypothetical reasonable individuals who have a particular response. The idea that such quantification is possible will almost always be an illusion, and illustrates how the test of "a not insignificant number" distracts from the terms of the statutory prohibition and the guidelines to its application laid down in Campomar.
(g) While, in our view, s 18 of the ACL (and analogous provisions) do not require the satisfaction of the “significant number” test in order to establish contravention, a party may choose to put its particular case that way. That is, it is open for a party to seek to establish that conduct is misleading by establishing that persons were in fact misled, and in such cases it may be necessary to establish that the number of such persons was significant in order to persuade the court that the conduct was misleading or deceptive or likely to mislead or deceive: see National Exchange at [23]. But that will be a function of how the case is put; it is not a requirement inherent in the statute.
193 As this comprehensive passage (relying on High Court authority) in TPG is correct, it is unnecessary to refer to additional authority. However and for the same reason, just as the application in TPG of the wrong test at first instance made no difference to the Full Court’s determination, it does not in this case either.
194 While dealing with TPG, it is convenient on a slightly different point to touch upon the question of qualifications to statements made in communications. In TPG, the Full Court said (at [25]):
A question that commonly arises is whether a publication or communication is misleading when it contains a misleading statement in one place but also contains another statement, perhaps in a different place in the publication or communication, which remedies the misleading character of the first statement. Ultimately, the question is one of overall assessment of the publication or communication. The correct approach was summarised by Edelman J in Australian Competition and Consumer Commission v Valve Corp (No 3) [2016] FCA 196; 337 ALR 647 (at [214]) (upheld in Valve Corp v Australian Competition and Consumer Commission (2017) 258 FCR 190, noting that there was no challenge to his Honour’s relevant statements of principle (recorded at [158]):
One consequence of the need to consider the conduct in light of all relevant circumstances is that any allegedly misleading representation must be read together with any qualifications and corrections to that statement. Hence, although a qualification to a statement might be effective to neutralise an otherwise misleading representation, this might not always be so, particularly if the misleading representation is prominent but the qualification (often linked to the representation by an asterisk) is not: Medical Benefits Fund of Australia Limited v Cassidy [2003] FCAFC 289; (2003) 135 FCR 1, 17 [37] (Stone J). As Keane JA expressed the point, the qualifications must have “the effect of erasing whatever is misleading in the conduct”: Downey v Carlson Hotels Asia Pacific Pty Ltd [2005] QCA 199 [83].
(Emphasis added.)
195 It is appropriate, therefore, to address the characteristics of ordinary and reasonable consumers in the manner asserted by Trivago. It is also however true, as the ACCC contends, that the list of attributes advanced by Trivago is incomplete and selective so as to support the Trivago case.
196 It is correct, as the ACCC contends, that to those characteristics one would add that ordinarily, reasonable users of search engines know that no humans intervene between the entry of a search query and a return of a list of results. Such users understand search results to be a ranked list that corresponds to the search terms or parameters that they have entered. They understand that search engines automatically rank results in an order that is likely to be optimal for the searcher. The first ranked result is, or is likely to be, the most relevant result with subsequent results progressively less relevant. Secondly, the fact that consumers “are able to take time”, as contended for Trivago, to study information does not accord with the conclusion, which his Honour was perfectly entitled to reach on the evidence adduced before him, that in reality consumers make quick decisions on the Trivago website. More will be said on this below.
197 Taking into account these additions to the knowledge base, any ordinary and reasonable consumer using this service would conclude that the Top Position Offer was the cheapest available offer conforming to the search terms or parameters entered by the consumer, or an offer other than the cheapest offer which held some other characteristic which would make it more attractive to that consumer than other available offers.
198 Trivago argues that the Top Position Representation is inconsistent with the various contextual matters which it emphasises, including the Our Recommendations information button hover-over and the link to the “Learn how Trivago works” page. It says that the ordinary and reasonable consumer can be taken to be familiar with these standard ways in which additional information can be obtained when searching for accommodation online and will take all of these contextual matters into account. It draws on observations in Australian Competition and Consumer Commission v Jetstar Airways Pty Ltd [2015] FCA 1263 per Foster J (at [33], [163], [164], [168] and [174]). Trivago argues that in finding to the contrary, the primary judge adopted an unrealistic and unsound approach to the operation of websites such as this as if they were fleeting advertisements that could not be closely considered. That information is delivered sequentially, and not as though it was all on one page at once, is an important and valid point made by Trivago. It is fair to say that the totality of information needs to be evaluated, but it may be that different degrees of importance are attached to the different sources of information based in part on their positioning in the context of the website as a whole.
199 Trivago complained about the primary judge’s “erroneous” rejection of Professor Parkes’ expert opinion that consumers looking for a hotel room are known to search extensively on multiple platforms before purchasing, referencing one study that placed the average hotel purchaser spending 191.4 minutes researching online and visiting 34 sites before making a purchase. Trivago complains that the primary judge preferred ‘his personal subjective opinion’ on this topic that it was unlikely that many consumers would spend this period of time or anything approaching it on searches involving the Trivago website.
200 Trivago acknowledges that the primary judge did additionally mention the evidence from Professor Slonim regarding “satisficers”, but seems to have overlooked the fact, Trivago says, that this evidence was based upon a false assumption that consumers would have to pay the full price of booking a room when they made the booking. Trivago notes that the primary judge queried whether the statements in the hover-over would be sufficient to dispel the impression created by the Top Position Offer because consumers may not place their mouse over the Our Recommendations information button and may not see the hover-over.
201 However, Trivago says that without answering this question, and without taking into account the fact that the adjacent Second, Third, and Fourth Position Offers were often equal to or lower than the Top Position Offer, the primary judge dismissed the hover-overs as being “opaque and insufficient” as they refer to “compensation paid to the booking site”, but did not explicitly state that compensation was being paid to Trivago. The primary judge also added that the word “compensation” does not really convey the nature of the CPC mechanism and did not “clearly disclose the significance of the CPC in the selection of the Top Position Offer”.
202 Trivago says this approach was misdirected. It was not necessary, Trivago argues, to disclose either the nature of the CPC mechanism or the significance of the CPC in the Trivago Algorithm in order to dispel the alleged Top Position Representation. Rather, the issue was whether the contextual statements would dispel any notion or potential implication that an ordinary and reasonable consumer might arrive at, that the Top Position Offers were the cheapest offers for that hotel or had some other characteristic which made them more attractive than any other offer for that hotel. The primary judge failed to grapple with the correct question, Trivago contends.
203 The essential argument for Trivago is that properly analysed, both elements of the alleged Top Position Representation, if they existed, were dispelled by the listed disclosures. Those disclosures convey that Top Position Offers were selected having regard to a range of matters which included price and compensation paid by Online Booking Sites. In this regard, and contrary to the primary judge’s finding, Trivago says the ordinary and reasonable consumer would understand that it was compensation paid to Trivago. Expressly stated in what Trivago describes as being prominent text “How payments to us affect ranking” as well as in the “Learn how Trivago works” page of the Trivago website, which stated: “We also take into account the compensation booking sites provide us with when a user clicks on an offer”.
204 Trivago says that taking all these matters into account, the ordinary and reasonable consumer would not have construed the Trivago website in the absolute sense of the Top Position Representation as contended for by the ACCC and found by the primary judge. Rather, such a consumer would have understood that the Top Position Offers were put forward as Trivago’s recommended offers, not because they were always the cheapest offers, as that was denied on the face of the website, but because in Trivago’s assessment they were the most attractive and relevant to consumers. Additional information that was readily accessible through the “Our Recommendations” information button and associated links explicitly confirmed this to be the case.
205 While the Court, on appeal, does not have the benefit of seeing the experts give evidence in person before the primary judge, it does have the same opportunity to evaluate the content of the Trivago website and advertising that was available to the primary judge.
206 On this issue, (but for the reference to a not insignificant number in Australian Competition and Consumer Commission v Get Qualified Australia Pty Ltd (in liq) (No 2) [2017] FCA 709), no exception was taken to the statements of principle set out by the primary judge (at [185]-[188]) to the following effect:
185 In relation to the first step, in Australian Competition and Consumer Commission v Jetstar Airways Pty Ltd [2015] FCA 1263, in the context of ss 18 and 29(1)(i) and (m) of the Australian Consumer Law, Foster J noted at [39] that where a headline representation is sought to be qualified by other material, the qualifying material must be sufficiently prominent to prevent the headline representation being misleading. The degree of prominence required will vary with the potential of the primary statement to be misleading (citing Medical Benefits Fund of Australia Ltd v Cassidy (2003) 135 FCR 1 at [37]-[41] per Stone J, Moore and Mansfield JJ agreeing). In this regard, it is the overall impression created by the representation that must be assessed.
186 The second step involves an objective determination by the Court of the effect of the representations on the ordinary or reasonable members of the class of consumers to whom the conduct is directed: Campomar at [102]; Google at [7] per French CJ, Crennan and Kiefel JJ. Conduct is misleading or deceptive, or likely to mislead or deceive, if it has a tendency to lead into error: Australian Competition and Consumer Commission v TPG Internet Pty Ltd (2013) 250 CLR 640 (TPG) at [39] per French CJ, Crennan, Bell and Keane JJ. As Allsop CJ held in Coles at [39], “[t]he causing of confusion or questioning is insufficient; it is necessary to establish that the ordinary or reasonable consumer is likely to be led into error.” As this passage indicates, and as the High Court held in Campomar at [105], the focus of the inquiry is on the meaning that would be conveyed to a hypothetical ordinary or reasonable member of the relevant class of consumers. In that context, it is relevant to consider: whether any of the alleged reactions are extreme or fanciful such that they should not be attributed to the ordinary or reasonable members of the class of consumers (Campomar at [105]; REA Group Limited v Fairfax Media Limited [2017] FCA 91 at [18] per Murphy J); and whether a not insignificant number of consumers are likely to have been led into error (Australian Competition and Consumer Commission v Get Qualified Australia Pty Ltd (in liq) (No 2) [2017] FCA 709 at [42] per Beach J; Australian Securities and Investments Commission v Westpac Banking Corporation (No 2) (2018) 266 FCR 147 at [2279] per Beach J).
187 Whether conduct in relation to a particular class of consumers is misleading or deceptive is a question of fact to be resolved by a consideration of the whole of the impugned conduct in the circumstances in which it occurred: Campbell v Backoffice Investments Pty Ltd (2009) 238 CLR 304 at [24]-[25] per French CJ, at [102] per Gummow, Hayne, Heydon and Kiefel JJ; TPG at [49]; Aldi Foods Pty Ltd v Moroccanoil Israel Ltd (2018) 261 FCR 301 at [74] per Perram J.
188 In TPG, French CJ, Crennan, Bell and Keane JJ observed at [50] that misleading or deceptive conduct may occur, not only when a contract has been concluded under the influence of a misleading advertisement, but also at the point where members of the target audience have been enticed into “the marketing web” by an erroneous belief engendered by an advertiser, even if the consumer may come to appreciate the true position before a transaction is concluded.
(Emphasis added.)
207 Having regard to the approach indicated by these cases, can it be said that any of the matters relied upon by Trivago negated a dominant message on its website to the effect that the Top Position Offers were the best offers for particular hotels either in terms of price or some other characteristic?
208 The features of the Trivago website, to the extent they may diminish either limb of the Top Position Representation were not adequate in the overall sense required, as discussed for example, in Australian Competition and Consumer Commission v Valve Corporation (No 3) (2016) 337 ALR 647; [2016] FCA 196 per Edelman J (at [214], [244]-[245] and [251]-[256]). The suggestion that the representation can be qualified by the hovering of a cursor over the information button, which will produce text was quite appropriately regarded by the primary judge as being insufficient to qualify the representation. From 6 August 2017, the Recommendations information button hover-over stated, amongst other things:
In determining the price to display in the leading position of our search results, we consider a variety of factors, including price, the likelihood that you will find your ideal hotel, your ability to complete a booking after you click on a search result and the level of compensation provided by the booking sites we cover. In order to make more information available on pricing options to our users, additional prices are listed in the “More deals” slide-out.
209 For the remainder of the Relevant Period, from 27 January 2018 until about October 2018, the text of the hover-over read:
The ranking results reflect your search criteria and our assessment of the attractiveness of the offer compared to other offers available on our site. It also reflects the compensation paid by the booking site. Learn how Trivago works.
(Emphasis in the original.)
210 Since about 17 April 2019, directly next to the Our Recommendations information button the Trivago website stated in prominent text “How payments to us affect ranking”. If hovered over, this entry would take one to the text extract in the previous paragraph. Since on or about 3 July 2018, further information was provided by a hyperlink to a “learn how Trivago works” page. Trivago argues that the alleged Top Position Representation is inconsistent with these contextual matters.
211 It is not possible to be satisfied that the Top Position Representation was dispelled, as Trivago contends, by any of the specific factors to which Trivago refers, namely, the listing of cheaper prices in the Second, Third and Fourth Position Offers (the frequency of which Trivago has exaggerated on appeal); the existence of the “More Deals” button; the description of results as a “default sort”; or the addition of attempted qualifications.
212 Contrary to Trivago’s submissions, it was not “often” the case that there were cheaper prices displayed, indeed, in many cases there were no prices displayed which were cheaper than the price of the Top Position Offer among the other prices listed in the initial offers. The statement of agreed facts and its annexures made this clear. It was frequently not the case that information displayed to consumers was capable of displacing the Top Position Representation. There were occasionally lower prices shown among the initial offers, but they were displayed less prominently and less attractively than the Top Position Offer.
213 The primary judge accepted the expert evidence, as he was entitled to do, in relation to “satisficers”, who were subject to “present” and “loss aversion” biases, who were unlikely to notice or attach significance to occasional lower prices displayed in this way. The primary judge expressly accepted, as his Honour was entitled to do, (at [158(g)]) that most people (including those referred to in the Professor Slonim’s report as “satisficers”) who face too many options, rather than proceeding to a more intense search, instead succumb to decision paralysis and accept the first choice that meets their inflexible attribute needs (location, dates, number of people) and is acceptable to (i.e., satisfies) their other preferences. For instance, if the price is within some range of acceptability (e.g., under $200 or under $300) and the room has two beds, then they will be satisfied with the choice and proceed to book the room. His Honour also accepted the reasons given for this behaviour as follows (at [158(g)]):
… First, once a consumer finds an initial accommodation that is satisfactory, any rooms that might be found with a further search are unlikely to be evaluated as better due to loss aversion given that there is a high likelihood that even if some attributes are better with the alternative room, other attributes are likely worse, and thus a loss on these attributes will receive greater weight in the evaluation. Secondly, consumers in many cases may not believe that they will be able to find a better deal, either because they do not think that they have the skill, or that it will take too much time to find a potentially better deal (present bias), or that there is not a better deal to be found if they search further. Moreover, in many decision-making situations, consumers often stay with defaults such as the first option they become aware of that is satisfactory (known as a status quo bias) because they lack confidence that another option might actually be a better deal.
214 It was entirely open to the primary judge to be satisfied as to this evidence. There is no reason to think that it is not plausible. His Honour also had the benefit of evaluating the evidence given viva voce.
215 In relation to the More Deals button, until 5 December 2017, that button did not display an offer price at all. So, again, it conveyed no information capable of displacing the Top Position Representation. After that date, it did display the cheapest priced hotel room offer from an Online Booking Site that had met Trivago’s minimum CPC threshold. In many cases, that price was not lower than the price of the Top Position Offer and so it was not capable of displacing any Top Position Representation.
216 Additionally, the More Deals offer would not necessarily be the lowest available price because if an Online Booking Site had made a CPC bid for an offer with a cheaper price than the Top Position Offer, but had not exceeded Trivago’s minimum CPC threshold, the offer was simply not displayed by Trivago at all. Trivago’s argument entirely glosses over this fundamental factor.
217 On appeal this Court also forms a view as to whether the initial qualifications added by Trivago to the website from October 2018 were effective. They were not. They were only visible to that subset of consumers who came across them by hovering a mouse cursor over the relevant text. Otherwise, there was nothing indicating even the existence of any hover-over text which might alert consumers to the fact that there was some important qualification to an otherwise attractively displayed price. Further, the primary judge’s characterisation of Trivago’s qualifications as “opaque and insufficient” (at [216]) were correct.
218 The qualifications mysteriously describe CPC bids as being the level of “compensation” provided by booking sites without identifying to whom the compensation was paid, or disclosing how they influenced the choice of offers to the consumer. In this context the word “compensation” is an ambiguous one which fails to state clearly the character of the CPC as a fee which booking sites paid to Trivago in order to be eligible to be displayed in Trivago’s search results. The qualifications set out on the “Learn how Trivago works” page (see [46(b)] above) did not advance this position. In each instance, the reference to the compensation Trivago would receive was listed at the end of the text, preceded by reference to the “variety of factors” or the “number of relevant factors” that are used to determine the Top Position Offer. Framed in this way, the qualification failed to disclose the significant role the CPC had in the calculation of the Top Position Offer, being at least the second most important factor on the evidence of Professor Parkes.
219 Trivago submits that it was not necessary for it to disclose either the nature of the CPC mechanism or the significance of the CPC in Trivago’s Algorithm. It was sufficient, it says, simply for the contextual statements contained in the hover-overs to dispel any notion or potential implication that an ordinary and reasonable consumer might have that the Top Position Offers were the cheapest offer for that hotel or had some characteristic which made them more attractive than any other offer for that hotel. Trivago says that the primary judge failed to address the correct question in this regard.
220 This submission cannot be accepted. The primary judge identified the correct question (at [216]) as being “whether any statement contained in the hover-over would be sufficient to dispel the impression created by the Top Position Offer …” (Emphasis added.). Given the nature of the Top Position Representation, and the role of the CPC in determining the Top Position Offer, it is difficult to see how the representation could have been dispelled without a more detailed disclosure of the CPC. This is particularly so because, as has been laboured in these reasons and by the primary judge, Trivago’s minimum CPC threshold excluded certain, and potentially cheaper or more attractive offers from even appearing on its website.
221 It is difficult to escape the conclusion that Trivago chose to use unclear language to avoid providing clear and accurate information to consumers as to the importance of the CPC mechanism in displaying offers.
222 The suggestion that consumers understood Trivago to be giving an “opinion” as to what it believed to be the cheapest or best offer, rather than conveying a message of objective fact is entirely unfounded. Clearly, users of a search engine expect that the results displayed to them have been ranked and arranged in response to their search inputs and other relevant attributes of the hotel room, such that an objective assessment of the available offers is presented. Consumers would not anticipate that Trivago was expressing an opinion as to which was the most suitable hotel. There was no human intervention as consumers would well understand. Consumers would certainly not understand that the Top Position Offer was determined to a significant degree (substantially understated by Trivago on appeal) by the value of the CPC bids made by Online Booking Sites. A consumer would simply never anticipate that the quantum of a CPC bid had anything to do with his or her search query, and as the primary judge correctly found in this instance (at [177]), the contention that the CPC acted as a proxy for the quality and attractiveness of an offer was entirely unpersuasive.
223 Trivago has also submitted that the primary judge was expressing his personal subjective opinion about the amount of time consumers were likely to spend on the Trivago website. Again this is not so. As indicated, the primary judge adopted expert evidence suggesting to the contrary, evidence which was entirely open to him to accept and entirely plausible. Indeed, the suggestion that consumers would spend hours booking accommodation and look at dozens of sites is not only contrary to common sense, but inconsistent with Trivago’s very advertising – its raison d'être – which offered consumers the ability to easily and “instantly” compare prices, thereby obviating the need to spend hours browsing from site to site.
224 In short, there was no reason at all for a consumer to anticipate that there was any correlation between the value of a CPC bid by an Online Booking Site and the attractiveness (or, for that matter, relevance) of an offer to a person conducting a search via the Trivago website. Not that consumers know anything at all about CPC bids, nor did the general references to “compensation” in the hover-overs provide any meaningful information about the significant role they played in determining the Top Position Offer so as to dispel the Top Position Representation.
225 The Top Position Representation, as pleaded, was established. These grounds of appeal cannot succeed.
8.2 Grounds 6-11 – Was The Top Position Representation Misleading?
226 There is no dispute that the Top Position Offers were not always the cheapest available offers. Indeed, Trivago stresses that there were occasions on which there were cheaper offers identified in the Second, Third and Fourth Position Offers and in the More Deals slide-out. This, Trivago says, disposes of any question as to whether the first element of the representation was made. As to the second (other attractive qualities), Trivago stresses that the ACCC bore the onus of proving that:
(a) Trivago did not select the Top Position Offers by reference to any other characteristic which made them the most attractive offer; and
(b) Trivago selected the Top Position Offers primarily by reference to the CPC payment that Trivago would receive from the Online Booking Site that submitted the offer if a consumer clicked on the offer.
(Emphasis added, see [57(b)] above).
227 The primary judge, of course, concluded that the Top Position Representation was false and misleading because in many cases it was not the cheapest available offer and did not have some other characteristics that made it more attractive than any other offer. Trivago does accept that the Top Position Offers were not always the cheapest available, but it contends that the ACCC failed to prove that the Top Position Representation was false or misleading because it did not establish its case as put in either (a) or (b) above.
228 Trivago says the following matters were common ground between the computer experts called for the parties (being, as noted, Mr Bajanov for ACCC and Professor Parkes for Trivago):
(a) the raw inputs into the Trivago Algorithm included a number of inputs in addition to the CPC, such as offer price, the minimum offer price of competing offers, the maximum offer price of competing offers, the click-through rate and the priority modifier;
(b) secondly, and based on those raw inputs, the Trivago Algorithm calculates a “composite score” for each offer made by an Online Booking Site with respect to a hotel listing. The offer with the highest composite score becomes the Top Position Offer for that hotel listing; and
(c) when Trivago applies its Algorithm to determine the composite score of each offer and awards the top position to the offer with the highest composite score, the primary, most important, or weightiest factor influencing the selection of the Top Position Offer was the offer price, not the CPC.
229 Trivago says this evidence was sufficient to defeat the ACCC case. Trivago did not “select” the Top Position Offers “primarily” by reference to CPC, there was no further step in the calculation that gave the CPC some additional priority. Hence, the ACCC did not make out its case, Trivago contends.
230 This submission misconceives ACCC’s case in two respects. Trivago’s argument proceeds on a basis that overlooks the first element of the Top Position Representation, being that the Top Position Offer was not the cheapest offer in some cases. References to “any other characteristic” that could make an offer attractive, and the determination of the Top Position Offer “primarily by reference” to the CPC, in the case as put by the ACCC, must be read in this context where the Top Position Offer is not the cheapest. Trivago’s proposition at [228(c)] above is therefore misguided. Secondly, it must be remembered that even on the evidence of Trivago’s expert, Professor Parkes, the CPC was the second most important factor on all three dates analysed, with a relative importance between 33.8% and 44.8%, followed by the priority modifier ranging between 4.0% and 13.3% (see [82] above). The other factors identified by the experts were found to have no bearing (full list above at [76]). The ACCC’s case was not confined in the manner articulated by Trivago. As noted above, the ACCC put its case on these bases in the FACS (at [12] and [13]):
12 Contrary to the Top Position Representation, the Top Position Offer was not always the cheapest available offer and Trivago did not select it by reference to any other characteristic which may have made it the most attractive offer. Trivago selected the Top Position Offer primarily by reference to the value of the CPC Trivago would receive from the advertiser who submitted the offer.
13 Trivago used an algorithm to determine the Top Position Offer. The algorithm placed a significant weighting on the value of an advertiser's CPC bid. By applying the algorithm, Trivago gave greater prominence to offers from advertisers who had submitted higher CPC bids than to cheaper offers at the same hotel from advertisers who submitted lower CPC bids. In addition, from at least May 2017 to at least October 2017, Trivago set a minimum CPC amount. If an advertiser submitted a CPC bid below this amount, the algorithm prevented the Trivago website from displaying the advertiser's offer, even if it was the cheapest price offer Trivago had received. The result was that often the Top Position Offer was not the cheapest price offer Trivago had received for the relevant hotel.
(Emphasis added.)
Trivago’s argument also ignores what was claimed at [13] of the FACS, that in any event, Trivago’s Algorithm placed a significant weighting on the value of an Online Booking Site’s CPC bid.
231 Trivago also takes exception with the primary judge’s treatment of Mr Bajanov’s Gradient Boosting Model analysis, contending that it erroneously influenced his reasoning and his ultimate conclusion as to misleading conduct.
232 Trivago observes, that the ACCC’s expert, Mr Bajanov, did not examine how the Algorithm selected the Top Position Offer. Rather, he analysed the results of the Algorithm’s selection, especially the Top Position Offer, in order to seek to measure the relative importance of each of the inputs within that sample. This process looked at the outcome of Trivago’s Algorithm, but did not use the source code of the Algorithm.
233 Trivago says that not only is the approach completely agnostic to the Algorithm actually used by Trivago to select the Top Position Offer, but it is also backward looking and presupposes that all of the outcomes produced by the Trivago Algorithm, including the identity of the Top Position Offer, are already known. Trivago argues that Mr Bajanov’s analysis was irrelevant to the issues in the proceeding. That analysis, which did not look at Trivago’s Algorithm, but rather looked merely at outcomes, was irrelevant to whether Trivago in fact selected the Top Position Offers in the manner alleged by the ACCC. Trivago says that Professor Parkes gave compelling evidence to this effect, that he was “by far the more highly qualified expert”, and his evidence “ought to have been accepted”.
234 The primary judge found (at [121]) that it was not necessary to form a view about this issue and that the Gradient Boosting Model analysis may be of assistance in determining whether the alleged representations were misleading or deceptive. This was erroneous according to Trivago. The primary judge was “obliged” to accept the conclusion arrived at jointly by the experts that:
(a) Trivago used the Algorithm to calculate the composite score;
(b) it always selected the offer with the highest composite score as the Top Position Offer; and
(c) the offer price was the most important factor in selecting the Top Position Offer.
Trivago says that the source of the primary judge’s error was that he failed to correctly identify the “only” question that needed to be answered to address the case pleaded and presented by the ACCC, namely, what relative importance the Algorithm attributed to its various input factors when Trivago used the Algorithm to select the Top Position Offer. Trivago says that the experts were agreed that the Gradient Boosting Model could not answer this question (see the primary judge’s reasons at [117]-[118]).
235 This submission must also be rejected as a distraction. Although the relative importance of the factors used in Trivago’s Algorithm was the subject of strongest disagreement between the experts, the disagreement was not as significant as Trivago contends. It has offered no cogent reason beyond assertion as to why inferences cannot be drawn as to the operation of the Algorithm based on the outcomes produced. Although they disagreed about the extent of the CPC’s role, Professor Parkes was also of the opinion that it played a determinative role in cases where the Top Position Offer was not the cheapest. As the primary judge noted (at [128]), Professor Parkes elaborated in his oral evidence:
When it is [the case] that the top position offer is not the cheapest offer, the variation between those offers that best explains that decision is the CPC. I agree with that. When they are priced differently, the variation that best explains it is coming through CPC. …
236 The experts did agree that the Gradient Boosting Model was relevant in determining “the relative importance that can be attributed to the input factors in explaining the listings outcomes observed” on the dates analysed by the experts: see the primary judge’s reasons at [117(c)].
237 The primary judge said that the CPC was a “very significant factor” influencing the Top Position Offer, indeed the second most important factor. This, however, Trivago says, was not the ACCC’s case. In addition, the proposition that the CPC was a “very significant factor” in the Trivago Algorithm did not provide an answer to the essential question of whether the Top Position Offers (where they were not the cheapest), were selected because they had some other characteristic that made them more attractive than any other offer for the hotel. Rather, Trivago argues, it begged the question of what other factors were in play. It also ignored the fact that the CPC was not directly used in the Algorithm, what was used was the priority CPC in the manner discussed above in these reasons (at [71]-]75]).
238 These submissions cannot be accepted. Trivago persistently ignores the ACCC’s case as presented (at [13]) in the FASC. It is convenient to set out again the FACS (at [12]-[13]):
12 Contrary to the Top Position Representation, the Top Position Offer was not always the cheapest available offer and Trivago did not select it by reference to any other characteristic which may have made it the most attractive offer. Trivago selected the Top Position Offer Position Offer primarily by reference to the value of the CPC Trivago would receive from the advertiser who submitted the offer.
13 Trivago used an algorithm to determine the Top Position Offer. The algorithm placed a significant weighting on the value of an advertiser's CPC bid. By applying the algorithm, Trivago gave greater prominence to offers from advertisers who had submitted higher CPC bids than to cheaper offers at the same hotel from advertisers who submitted lower CPC bids. In addition, from at least May 2017 to at least October 2017, Trivago set a minimum CPC amount. If an advertiser submitted a CPC bid below this amount, the algorithm prevented the Trivago website from displaying the advertiser's offer, even if it was the cheapest price offer Trivago had received. The result was that often the Top Position Offer was not the cheapest price offer Trivago had received for the relevant hotel.
(Emphasis added.)
Trivago’s contentions are premised on a misstatement of the ACCC’s case. Trivago have focussed only on the words “selected” and “primarily” where they appear in the relief sought in the Further Amended Originating Application and in the final sentence of the FASC. It has overlooked the articulation of the ACCC case in the FACS at [12] and the first sentence in [13].
239 It is quite clear that the ACCC’s case included the proposition that the Top Position Representation was false and misleading because Trivago’s Algorithm “placed a significant weighting on the value of an advertiser’s CPC bid”. As the ACCC contends, Trivago’s argument would not even be a good pleading point, given the way the parties conducted their case, but certainly in a case which proceeded by exchange of concise statements and evidence, the point goes nowhere.
240 The evidence was the same, whether in support of a contention that CPC bids were the “primary” driver of the Top Position Offer or that Trivago’s Algorithm “placed a significant weighting” on CPC bids. There can be no doubt that the latter proposition was established on the evidence.
241 But it should also be noted that the primary judge did not grant declaratory relief in the form sought by the ACCC. His Honour’s finding was that in many cases the Top Position Offer was not the cheapest offer for the hotel, nor did it have some other characteristic that made it more attractive than any other offer for the hotel. It was not necessary, in demonstrating those contraventions, for the ACCC to also prove that Trivago “selected” the Top Position Offers primarily by reference to the CPC payment that Trivago would receive.
242 The ACCC established the case articulated at [13] of the FACS, namely, that Top Position Offers were determined by Trivago’s Algorithm, which “placed a significant weighting on the value” of an Online Booking Site’s CPC bid, and filtered out offers which did not meet a minimum gain threshold. It is clear on the evidence that the experts were agreed on the following matters:
(a) CPC bids are an important component of the Algorithm which selects the Top Position Offer;
(b) the characteristics of an offer, other than price, are not a factor in the Algorithm;
(c) in a large number of cases, offer prices were clustered and the CPC bids were the primary factor driving the selection of the Top Position Offer;
(d) when a higher priced offer was selected as the Top Position Offer over the lower priced offer, the selection was attributed to a difference in the CPC bid more than any other factor; and
(e) the Algorithm filtered out offers which did not meet a minimum gain threshold or the confidential minimum CPC bid amount (see the primary judge’s reasons at [122], [125]-[126] and [221]).
243 There is no doubt, in light of those conclusions, that the primary judge was entitled to be satisfied that CPC bids were “a very significant factor in the selection of the Top Position Offer”: (at [204], [221] and [225]).
244 In the alternative, Trivago advanced an argument at first instance, and strenuously on appeal, that the CPC bidding mechanism acted as a proxy for the attractiveness of an offer and through it, that the Algorithm accounted for non-price attributes in calculating the Top Position Offer. The primary judge had found (at [177]), as he was entitled, that the CPC bids were certainly not a proxy for the quality and attractiveness of an offer. Trivago’s contentions, it will be recalled, were based almost exclusively on the evidence of its expert economist, Mr Smith. Trivago chose to call this evidence; it did not have to.
245 Trivago stresses that the parties’ experts agreed that the CPC was inputted into the Algorithm by multiplying it with the priority modifier which was strongly correlated with, or heavily dependent on the Landing Page Score. Trivago says that through the combination of the CPC with these factors, the Algorithm took into account non-price attributes and the attractiveness of an offer in ways additional to the price. The primary judge rejected this submission, holding (at [103]) that these factors only provided information about the quality of a Booking Site’s webpage, rather than the quality of the actual accommodation offer to the consumer. Trivago says this distinction is without substance, but that must be rejected. The reasoning and conclusion of the primary judge was entirely correct. Trivago’s business records describe this as being a measure of the quality of the Online Booking Sites’ websites and their ease of usage. How the attractiveness of its booking site bears any relevance to the attractiveness of the actual accommodation booked by a consumer has never been explained with any degree of precision or logic.
246 His Honour held (at [177]) that “there is no reason to presume that an Online Booking Site that expects a higher gain necessarily, or even ordinarily, has a higher quality offer”. Trivago says this conclusion is illogical, especially given the “undisputed economic evidence” that the primary judge otherwise accepted. The evidence from Mr Smith was that as a matter of economic principle, an Online Booking Site with a higher quality hotel offer is likely to offer a higher CPC bid to Trivago. This was said to be because an attractive offer is more likely to be converted into an actual booking once the consumer clicks through to the Online Booking Site. Trivago says that, because as a matter of “basic economics” a higher CPC bid reflects a higher anticipated conversion rate, the CPC mechanism implicitly accounts for non-price attributes and the overall attractiveness of an offer in calculating the Top Position Offer.
247 The primary judge accepted that a higher CPC meant a higher expected gain per click, but not that it depended on the higher conversion rate that was linked to a higher quality offering, which Trivago says “makes no sense”. Trivago also complains that the primary judge erred in dismissing Professor Parkes’ empirical analysis which showed that Top Position Offers had superior non-price features than lower priced offers such that the use of the priority CPC must be conveying information about the quality of the offers. The point is said to follow as a matter of “rational economic analysis” supported by statistically significant data.
248 These submissions are at best another distraction. The primary judge was entitled, and quite correct, to give little weight to Mr Smith’s opinion that CPC bids were in some way a proxy for factors relating to the quality or attractiveness of the hotel room offer from the perspective of a consumer. His Honour rightly observed (at [177]) that there was no reason to assume that an Online Booking Site that expects a higher gain necessarily or even ordinarily has a higher quality offer to the consumer. Significantly, Mr Smith did not undertake any analysis using actual data to see whether higher CPC bids were associated with higher quality websites or offers, he had not examined Trivago’s Algorithm, and had not differentiated between the quality of an Online Booking Site’s website and its offers. At its highest this evidence was purely theoretical. It was entirely understandable that the primary judge would prefer the evidence of Mr Bajanov that because there was no actual evidence in relation to how Online Booking Sites estimate or predict their prospective conversion rates in deciding a CPC bid, it cannot be inferred that the value of the bid conveys any information about the quality of the underlying hotel room offer.
249 Trivago attempted to demonstrate that there were, if not superior price features about the Top Position Representation, then some other attractive qualities. Nothing in the features described by the experts called for Trivago demonstrated that any such qualities determined the offer made in the Top Position Representation. It is not the case that there was a reversal of the onus of proof requiring Trivago to prove that there were such factors. Trivago chose to do so. It is simply that the evidence it called was not accepted. The operation of the Algorithm was a matter peculiarly within the knowledge of Trivago, but it chose not to call any of its officers or employees to give evidence. It did not, as the ACCC observed, even provide its own expert, Professor Parkes, with all the information necessary to enable him to analyse fully the operation of the Algorithm. None of the input factors identified by the experts as informing the determination of the Top Position Offer had any apparent relevance to the quality of the hotel room offering. That was a sufficient foundation, on the expert evidence, for the primary judge’s finding. The ACCC bore the onus of proving its allegations. Trivago sought to suggest otherwise by its expert evidence. It was not compelled to do so. It was a matter of its forensic choice.
250 Trivago was the party who advanced the assertion that CPC bids were a proxy for the quality or attractiveness of hotel room offers. Each party adduced evidence in relation to that assertion. As the ACCC observed, even within Professor Parkes’ limited sample, there were numerous occasions when Top Position Offers had one or more inferior non-price features than lower priced offers. The primary judge preferred the expert evidence adduced for the ACCC. His Honour did so on an entirely conventional and sound basis.
251 The ACCC established that the Top Position Representation was misleading. These grounds of appeal are not made out.
8.3 Grounds 12 And 13 – Was The Red Price Representation Made?
252 Trivago contends that contrary to the primary judge’s finding, the Red Price Representation would not be conveyed to the ordinary and reasonable consumer. When the Red Price appeared on the Initial Search Results Page during the third relevant sub-period, a hover-over would be displayed stating, that “… this is the cheapest deal from the most expensive booking site with offers for this hotel. It may relate to a room type or amenities package that differs from other offers on our site”. The primary judge, however, dismissed the hover-over on the basis that it was confusing and did not dispel the Red Price Representation (at [257]-[259]). Trivago complains that the primary judge did not explain these matters. Trivago says the hover-over plainly dispelled the alleged representation. It stated expressly that the Red Price may relate to a different room type in the hotel.
253 Trivago contends that ordinary and reasonable consumers understand that further information can be obtained by clicking on or hovering over items, such as the Our Recommendations information button, the Red Price or the More Deals button in the fashion with which such consumers are familiar.
254 Trivago correctly contends that the statutory provisions should not be given a meaning that is impractical and unworkable in the context of internet searches. Trivago is also correct in saying that information must be provided progressively by means of links or hover-overs because not all information can be displayed on a screen at one time. Additionally, Trivago says that the documents relied on by the ACCC as giving rise to the Red Price Representation make it clear that when the Red Price is clicked through, an offer is identified which differs in marginal respects from the Top Position Offer. However, of the two, the Top Position Offer is superior to the Red Price, not inferior. This illustrates that the alleged representation is inherently vague when it uses the concept of the “same room category” and was not, therefore, established by the ACCC.
255 The primary judge found that the Red Price Representation was that the Red Price provided a comparison with the Top Position Offer of prices offered for the same room category in the same hotel.
256 There was no complaint by Trivago that if the Representation was made (which it denies), it was false or misleading. The Red Price was not necessarily a “like-for-like” comparison with the Top Position Offer price. Trivago accepts that the Red Price could relate to a different room type in the same hotel or a room with superior amenities.
257 This is very much a matter of impression, but it has not been shown that the primary judge erred. It was open to the primary judge to conclude, as his Honour did (at [256]), that the natural inference an ordinary and reasonable consumer using Trivago’s website would draw would be that the Top Position Offer and the Red Price were being presented as comparable offers, apart from price. In other words, the offers were, apart from price, “like-for-like”. There would be little utility in the comparison if it simply showed that a room with superior amenity was more expensive than one with inferior amenity.
258 The primary judge was clearly not satisfied that the text displayed to consumers who did choose to hover their mouse cursor over the Red Price was sufficient to dispel the Red Price Representation. In order to obtain the benefit of the explanation for which Trivago contends, a consumer would have to, in the first place, be aware of the hover-over text. The primary judge was not satisfied that that would be so in many cases.
259 Trivago did not attempt to alert customers to the fact that the natural inference one might draw from seeing a higher price in red text over a lower price in green was, or might be, erroneous by the use of a conventional visual cue, such as a clear asterisk or footnote.
260 In any event, we agree with the primary judge (at [258]) that the content of the hover-over text was confusing. The meaning of the expression “the cheapest deal from the most expensive booking site with offers for this hotel” is not straight forward language understood by ordinary and reasonable consumers. Rather than simply telling the consumers that the Red Price may not relate to the same room type or have the same benefits as the room to which the Top Position Offer price related, Trivago confusingly stated that the characteristics of the Red Price may differ from “other offers on our site”.
261 Generally speaking, Trivago chose to communicate such matters in ways which were not simple or clear. The suggestion by Trivago that the words “the same room category” in the ACCC’s formulation of the Red Price Representation are vague in themselves, cannot be accepted.
262 “The same room category” are the very sorts of clear words which Trivago avoided using. There may well be reasonable consumer confusion between, say, “deluxe” and a “superior” room, but there would not be between a superior room and another superior room. Such rooms would be expected to be rooms in the same category. Further Trivago itself had admitted that the Red Price could relate to a different room type in the same hotel.
263 These grounds of appeal are not made out.
8.4 Grounds 14 And 15 – Was there an Additional Conduct Representation?
264 The primary judge concluded that the Trivago website did not provide an impartial, objective and transparent price comparison service and that it did not quickly and easily identify the cheapest rates available for a hotel room in response to a consumer’s search. Trivago contends that neither of these propositions is capable of being sustained. Both the Second, Third and Fourth Position Offers, and the More Deals button identified for consumers the existence of offers which were equal or lower when there were equal or lower prices than the Top Position Offer available. In addition, the Trivago website did not direct users to offers made by Online Booking Sites which made higher CPC bids, irrespective of the price offered to consumers. The weightiest and most influential factor in the choice of the Top Position Offer under the Algorithm, Trivago contends, was the offer price, not the CPC. The disclosure on the face of the website, Trivago says, made it clear that it was providing its recommendation as to the Top Position Offer, taking into account a range of factors. It was not purporting simply to identify the cheapest available offer. Trivago says the content on its website could not lead ordinary and reasonable consumers to believe that Trivago was always presenting the cheapest available offer.
265 Although this has been described as an Additional Conduct Representation, it is in sense a wrapping up of all the factors. The finding by the primary judge was that by making the Top Position Offers, the Top Position Representation, the Strike-Through Representation and its advertising on television prior to 2 July 2018, Trivago conveyed the Additional Conduct Representation to consumers, that is, that its website provided an impartial, objective and transparent price comparison which would enable them to quickly and easily identify the cheapest available offer for a particular (or the exact same) room at a particular hotel. For Trivago to succeed on grounds 14 and 15, it must succeed in establishing that the Top Position Representation was not conveyed. The primary judge was correct in finding that the Top Position Representation was conveyed.
266 Trivago depends on the display of the other initial offers and the More Deals button to dispel the Additional Conduct Representation, but the reasons why they do not achieve this have already been addressed. In many cases, there were no cheaper prices than the price of the Top Position Offer among the other prices listed in the initial offers and no information was displayed to consumers capable of displacing the Top Position Representation. Further, where lower prices were displayed, they were less prominent than the Top Position Offer. The finding of the primary judge was that users of the Trivago website included satisficers who were subject to “present” and “loss aversion” biases and were unlikely to attach significance to the other offers. There was no visual cue to qualify the Top Position Offer in any way except for certain hover-overs which it has been reasonably held would not have been obvious or apparent to all consumers. As already noted, the More Deals button did not display an offer price at all until 5 December 2017. After that date, it displayed the cheapest priced hotel room offer from an Online Booking Site that had met Trivago’s minimum CPC threshold. In many cases that price was not lower than the price of the Top Position Offer and so was not capable of displacing the Top Position Representation. In those other cases where the price was lower, it does not follow that it was the lowest available price because if an Online Booking Site had made a CPC bid for an offer with a cheaper price than the Top Position Offer, but had not exceeded Trivago’s minimum CPC threshold, the offer would not have been displayed by Trivago at all.
267 The relevant advertising was accurately summarised by the primary judge (at [224]) where his Honour referred to the television advertisement set out above (at [28]), in which the presenter referred to the fact that there are so many different prices out there (i.e. online) “for the exact same room”. She said that Trivago “does the work for you and instantly compares the prices of over 600,000 hotels from over 200 different websites”. It was said that, instead of “searching for hours”, Trivago “makes it easy for you” to find the ideal hotel for the best price. The presenter said that you just go to Trivago, type in where you want to go, and “with two clicks” select your check-in and check-out dates and search; “It’s that simple”. It was said that Trivago shows you all the different prices “for the exact same room”. The overall impression was that Trivago was impartial, objective and transparent. Although not as lengthy and detailed, the other television advertisements were to similar effect. The Strike-Through Representation and the Top Position Representation were consistent with the messages conveyed in the television advertisements. While changes were made to the Trivago website during the course of the first relevant sub-period, they were not sufficient to dispel the impression created by the television advertisements.
268 This advertising, together with the way in which the Trivago website operated, clearly makes good the Additional Conduct Representation.
269 These grounds of appeal also fail.
8.5 New ground – were inferences wrongly drawn?
270 On the day before the hearing of the appeal, Trivago sought leave to amend the notice of appeal to add a ground of appeal addressing an inference twice drawn by the primary judge that the Red Price did not always relate to the same room category as the Top Position Offer.
271 The proposed new ground of appeal is:
The trial judge erred in finding that the Red Price Representation was misleading by reason of the inference he drew (but ought not have drawn) at J[260] and [272].
272 The ground relates to [260] and [272] of the primary judge’s reasons, which concerned findings as to the Red Price Representation in the third and fourth relevant sub-periods and were in these terms:
260 In making the Red Price Representation during the third relevant sub-period, Trivago engaged in conduct that was misleading and deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. It can be inferred from the second sentence of the hover-over that the Red Price did not always relate to the same room category as the Top Position Offer. There is no evidence from any officer or employee of Trivago to the contrary.
…
272 In making the Red Price Representation during the fourth relevant sub-period, Trivago engaged in conduct that was misleading or deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. Again, it can be inferred from the second sentence of the hover-over in relation to the Red Price (see [257] above) that the Red Price did not always relate to the same room category as the Top Position Offer. Again, there is no evidence from any officer or employee of Trivago to the contrary.
(Emphasis added.)
273 It will be recalled that the ACCC alleged that Trivago represented that the Red Price was a comparison with the Top Position Offer between prices offered for the same room category in the same hotel.
274 The context of the discussion in the primary judge’s reasons was as follows:
Red Price Representation
…
256. The way in which the Red Price was depicted during the third relevant sub-period can be seen from the screenshot set out in [74] above. The Red Price appeared in red text in the same column as the Top Position Offer. No other offers were set out in that column of the listing for the hotel. Although the Red Price was a smaller size than the Top Position Offer, the presentation of both offers in the same column of the listing conveyed that the two offers were comparable offers apart from price; in other words, that the offers were, apart from price, ‘like for like’. This is the natural inference from the way the two offers were presented. It is also supported by Professor Slonim’s evidence, albeit generally expressed in terms of the Strike-Through Price.
257. Throughout the third relevant sub-period, if a consumer’s mouse cursor hovered over the Red Price, the following text was displayed:
For your stay dates, this is the cheapest deal from the most expensive booking site with offers for this hotel. It may relate to a room type or amenities package that differs from other offers on our site.
258. Again, there is a question whether any statement contained in the hover-over would be sufficient to dispel the impression that the Red Price and the Top Position Offer were comparable offers (apart from price). Further, in my view the wording of the hover-over is confusing, and does not dispel the impression that would otherwise be given that the Red Price and the Top Position Offer were comparable offers (apart from price).
259. In light of the above, and notwithstanding the hover-over, I consider that Trivago represented that the Red Price was a comparison between prices offered for the same room category in the same hotel.
260. In making the Red Price Representation during the third relevant sub-period, Trivago engaged in conduct that was misleading and deceptive or likely to mislead or deceive, and thus contravened s 18 of the Australian Consumer Law. It can be inferred from the second sentence of the hover-over that the Red Price did not always relate to the same room category as the Top Position Offer. There is no evidence from any officer or employee of Trivago to the contrary.
275 The primary judge’s reasoning in the paragraphs leading up to the inference drawn at [272] are of a similar character.
276 Ultimately, leave to amend was not opposed and the amendment was allowed. The argument, however, fails for the reasons below. The words speak for themselves. They are Trivago’s own words in the hover-over which, by themselves, prove that the Red Price Representation was not correct.
277 Trivago contends that, as a matter of legal principle, the inferences drawn by the primary judge were not open and to draw such inferences was inconsistent with cases such as Luxton v Vines (1952) 85 CLR 352 (at 358), where Dixon, Fullagar and Kitto JJ said:
In a civil case where direct proof is not available it is enough if the circumstances appearing in evidence to give rise to a reasonable and definite inference. They must do more than give rise to conflicting inferences of equal degrees of probability so that the choice between them is a mere matter of conjecture.
See also Holloway v McFeeters (1956) 94 CLR 470 (at 480-481).
278 Trivago’s argument appears to be premised on the proposition that although Trivago informed consumers of the matters in the second sentence of the hover-over, the ACCC did not prove as a fact that what Trivago told consumers was true. Trivago’s argument is also founded on the assumption that the second sentences in [260] and [272] can be read in isolation from everything else which had happened in this factually complex case. It ignores the significance of the admissions made by Trivago in relation to the Strike-Through Price. The primary judge (at [34]) recorded admissions made by Trivago in its Concise Response prior to the commencement of the trial including the admission that the Strike-Through Price does not always relate to the same room category as the Top Position Offer. Trivago accepted that it followed from this, that some consumers were misled and that therefore there had been contraventions. Similarly, Trivago noted that from 20 November 2018, when the Strike-Through Price was replaced by the Red Price, the hover-over was also amended to include an additional statement that the price may relate to a room type or amenities package that differed from other offers on the site. This was the hover-over text which Trivago relied upon in relation to the Red Price in order to argue that it dispelled the admittedly misleading representation previously conveyed by the Strike-Through Price. The primary judge correctly held (at [256]-[259]) that this additional statement in the hover-over failed to dispel the impression created by the Red Price Representation.
279 It is clear therefore, that there was more on which his Honour relied in reaching this inference than Trivago suggests when regard is had to the context of the reasons. The inference was drawn against the background of the admission, both of fact and law, made by Trivago to the effect that the Strike-Through Price was false because, as a matter of fact, the Strike-Through Price did not always relate to the same room category as the Top Position Offer. Against that background, his Honour had the admission to take into account when he drew the inference in relation to the Red Price hover-over because, save for the inclusion of the additional statement from 20 November 2018 which his Honour dealt with (at [256]-[259]), exactly the same words had appeared in relation to the Strike-Though hover-over which had failed to dispel the impression created.
280 The findings by the primary judge (at [208]-[209] and [229]-[232]) that qualifications to the Strike-Through Price in the hover-over from 29 March 2018 failed to dispel the impression created are not challenged by Trivago on this appeal.
281 These findings, together with Trivago’s admissions with respect to the Strike-Through Price, were essential context to his Honour’s finding with respect to the second sentence added to the hover-over when the Red Price was introduced in November 2018. Trivago had already said that it was not always engaging in a “like-for-like” comparison with respect to the Strike-Through Price. As his Honour correctly held, none of Trivago’s subsequent actions dispelled the impression created by the Strike-Through and Red Price Representations. The inference was not drawn in a vacuum.
282 The primary judge was, therefore, entitled to draw the inference that he did at [260] and [272] of the judgment.
283 This fresh ground also fails.
284 The appeal must be dismissed with costs.
I certify that the preceding two hundred and eighty-four (284) numbered paragraphs are a true copy of the Reasons for Judgment herein of the Honourable Justices Middleton, McKerracher and Jackson. |
