How Does Google Make Money: Search Ads and Text Ads Explained

C. Advertising, Including Search And Text Ads

84. The primary objective of all advertisers is to sell their product or service. Tr. 1371:12–24 (Dischler (Google)). Advertising is used to capture the attention of consumers, educate them, and drive them forward to a conversion, which for a product or service is a purchase. Tr. 3814:25–3815:13 (Lowcock (IPG)).

85. Google monetizes its general search services with the sale of Search Ads, which provides most of Google’s revenues. Tr. 7345:8–19 (Raghavan (Google)); Tr. 184:11–12 (Varian (Google)); UPX8065 at -773 (2021 total revenues of $258 billion and “Google Search & other” revenues of $149 billion); UPX7002.A (2021 “Search+” revenues of $146 billion and operating profit of $[redacted] billion); Tr. 1137:2–14 (Dischler (Google)) (Approximately two-thirds of Google’s revenues are from Search Ads.).

86. Online or digital advertising (including Search Ads) spending dominates the multi-billion-dollar U.S. advertising industry. For example, in 2021 alone, advertisers spent an estimated $316 billion on advertising across all media in the United States. Approximately onethird of this spend was on traditional (non-digital/offline) advertising, while the remaining majority of spend—approximately $211 billion—was on digital advertising. Tr. 5377:11–19 (Jerath (Pls. Expert)) (discussing UPXD103 at 3 (Total of amounts in 2021 U.S. “Total Media Ad Spending, by Media” chart is $315.79 billion.)).

Digital advertising includes ads on digital properties, such as on search engines, online websites, social media platforms, connected television, and other digital channels. Tr. 3803:9–16 (Lowcock (IPG)).

87. Traditional advertising includes television, print (e.g., newspaper and magazines), radio, and “out of home” (e.g., billboards). Tr. 5377:20–5378:5 (Jerath (Pls. Expert)) (discussing UPXD103 at 3).

1. Overview Of Digital Advertising

88. By ad spend, the two largest digital ad channels are display ads and Search Ads (including Text Ads). Tr. 5378:6–14 (Jerath (Pls. Expert)) (discussing UPXD103 at 3); UPX0006 at -329 (Oct. 2019 Google “Search State of the Union”). In 2021, U.S. Search Ad spend was approximately $86 billion. Tr. 5378:6–14 (Jerath (Pls. Expert)) (discussing UPXD103 at 3); DX0407 at -633 (MAGNA 2021 report predicted U.S. Search Ad revenue growth of 23.2% from $70 billion in 2020, which is approximately $86 billion in 2021.). Google sells both Search Ads and display ads.

a) Search Advertising

89. Search advertising (Search Ads) includes any ad that is served on a search engine results page in response to a consumer’s real-time search query. Tr. 3803:21–3804:15 (Lowcock (IPG)) (“Search advertising can be defined as advertising that you buy [that is shown] in response to people conducting a search on a search engine or platform.”); Tr. 3989:23–3990:5 (Juda (Google)); Tr. 5392:3–5393:9 (Jerath (Pls. Expert)) (Search Ads are “anything that is in response to a query.”); Des. Tr. 20:24–21:13 (Jain (Google) Dep.); Des. Tr. 54:15–20 (Ramalingam (Yahoo) Dep.). Search Ads can appear on both GSEs and on SVPs. Tr. 3853:3–15, 3925:24–3926:5 (Lowcock (IPG)); Tr. 5379:21–5380:14 (discussing Search Ads shown on UPXD103 at 4–5).

90. On a general search engine SERP, Search Ads are not shown in response to all queries, but instead on a subset. UPX0010 at -053 & n.6 (As of January 2020, “the vast majority of queries[–about 80%–]don’t show any ads at all.”); Tr. 8396:16–8398:17 (Israel (Def. Expert)). Search Ads on a GSE appear on SERPs in response to queries showing commercial intent, often when a user is contemplating purchasing a product or service. Des. Tr. 110:10–24 (Fox (Google) Dep.) (A commercial query is one where a user is “looking to conduct . . . a transaction for a product or service.”); Tr. 3686:22–3687:8 (Ramaswamy (Neeva)) (Queries for products or services have “strong commercial intent,” and political queries are also “commercially-oriented” before an election.); Tr. 398:13–399:2 (Varian (Google)).

91. GSEs that show Search Ads in the United States include independent GSEs such as Google, Bing, and Brave, as well as syndicated GSEs such as Yahoo, DuckDuckGo, Ask.com, and You.com. Tr. 5230:12–19, 5231:8–20 (Dijk (Booking.com)) (naming Google, Bing, and DuckDuckGo); Tr. 5125:19–20, 5126:1–13 (Booth (The Home Depot)) (naming Google, Bing, and adMarketplace); Tr. 6565:13–19 (Hurst (Expedia)) (“The only real comparable for the text ads [on Google] would be Bing.”); Des. Tr. 21:4–12 (van der Kooi (Microsoft) Dep.) (“The search advertising marketplace is a marketplace where advertisers participate by buying specific search ads for generic search engines like Google and Microsoft Bing.”); UPX0913 at -017 (Dr. Varian listing Google, Bing, and DuckDuckGo as ad-supported “general purpose” search engines).

92. Specialized search engines that show Search Ads in the United States include Amazon.com, travel websites like Expedia.com, app stores like Apple’s App Store, and Google Maps. Tr. 5380:9–14 (Jerath (Pls. Expert)) (discussing UPXD103 at 5 (illustrating Search Ads on Amazon.com and Expedia.com)); Tr. 5117:1–5 (Booth (The Home Depot)) (identifying Apple’s App Store); PSX00340 at -179 (discussing Search Ads on Google Maps); UPX0334 at -085 (identifying “special purpose search engines”: Amazon.com, Apple (also offers maps), Google, Facebook, and Microsoft); UPX0913 at -017 (Dr. Varian listing Amazon, eBay, Yelp, Travelocity, TripAdvisor as ad-supported “special purpose” search engines).

93. Search Ads includes Text Ads, infra ¶¶ 102–107 (§ III.C.1.b), ¶¶ § III.C.1.b, as well as shopping ads, hotel ads, travel ads, Amazon sponsored product ads, map ads, and app store ads. Tr. 5442:17–25 (Jerath (Pls. Expert)); Tr. 1177:2–4 (Dischler (Google)) (Text Ads and shopping ads are Search Ads.); Tr. 3922:10–3923:6 (Lowcock (IPG)) (Search Ads include Text Ads and shopping, local, travel, and vertical search ads.); Tr. 5236:11–13 (Dijk (Booking.com)) (Text Ads and hotel ads are Search Ads.); Tr. 5113:12–21, 5117:1–5, 5120:3–11, 5125:5–18 (Booth (The Home Depot)) (Paid search includes Text Ads, shopping ads, and app store ads.); UPX0032 at -144–50, -164–70 (depicting Text Ads, shopping ads, app ads, hotel ads, and map ads).

94. Figure 2 below, excerpted from a Google document, shows the components of Google’s mobile SERP and identifies the query, Text Ads, shopping ads, and organic results. UPX0012 at .004; Tr. 1182:3–5 (Dischler (Google)) (UPX0012 at .004 depicts three shopping ads next to a Text Ad.). Other examples, including of desktop SERPs, also appeared at trial. Tr. 5379:21–5380:5, 5380:9–14 (Jerath (Pls. Expert)) (UPXD103 at 4); UPX0012 at .003. For an example of a desktop SERP, see Figure 1, supra ¶ 60.

95. Shopping ads on google.com were formerly called Product Listing Ads, or PLAs, and the term PLA continues to be used interchangeably with shopping ad. Tr. 407:5–13 (Varian (Google)); Tr. 1181:16–24 (Dischler (Google)) (Shopping ads and PLAs are synonymous.); UPX0915 at -063 (Shopping ads were formerly known as (“f.k.a.”) “Product Listing Ads.”). As reflected in Figure 2, shopping ads—also know as product listing ads (PLAs)—are visual ads specific to an Individual physical product.

Shopping ads feature the image, name, price, and short description of the product; link directly to a page where the user can purchase the product; and are targeted using a product feed. UPX6032 at -652 (Google admission (Resp. 3) that shopping ads “typically include an image, a price, a title, and a store name”); Tr. 407:15–408:3, 424:1–3 (Varian (Google)) (Shopping ads or PLAs show one product per ad, include a picture and a price, identify the seller, and take users to the advertiser’s website); UPX0032 at -145 (depicting shopping ads); Des. Tr. 17:18–28:2 (McAteer (Google) Dep.) (PLAs can only display a single product).

96. The shopping ads carousel on google.com can include shopping ads for products from different sellers. Tr. 1350:12–18 (Dischler (Google)); UPX0032 at -145; UPX0012 at .004 (depicting shopping ads for same product from different sellers) (shown in Figure 2, supra ¶ 94).

97. Similar to shopping ads are Amazon sponsored product ads, which appear on amazon.com. Tr. 5379:21–5380:5, 5380:9–14, 5437:17–5438:4 (Jerath (Pls. Expert)) (“When you think of Amazon, shopping ads are sponsored product ads”; discussing UPXD103 at 4–5); Des. Tr. 141:4–7 (Jain (Google) Dep.) (Searches that give rise to PLAs are comparable to Amazon ads.).

98. Hotel ads or travel ads appear on google.com and travel websites like Expedia. Tr. 5380:9–14 (Jerath (Pls. Expert)) (discussing UPXD103 at 5); Tr. 9195:8–21 (Holden (Google)); Tr. 5232:2–11 (Dijk (Booking.com)) (Google hotel ads are a price comparison tool that Google puts into the paid search results.); Des. Tr. 36:25–37:9 (Silverman (Google) Dep.)

(Hotel ads include booking links, promoted hotels, and hotels commercial unit.); UPX0032 at -164–65, -170 (depicting Google hotel ads). Hotel ads are served using a product feed—like a PLA—that includes all of the rate and occupancy information for a particular brand and location. Des. Tr. 66:25–67:15 (Alberts (Dentsu) Dep.).

99. Map ads include information such as the advertiser’s address, customer rating, and hours of operation. UPX0032 at -166–67 (depicting map ads on google.com and the Google maps immersive).

100. App store ads can appear on google.com, Apple’s App Store, or on Google Play. UPX0032 at -148 (example of Play Store ad). For example, The Home Depot purchases Search Ads on Apple’s App Store to “prompt people to download The Home Depot app.” Tr. 5113:12– 16, 5115:22–5116:2, 5117:1–5 (Booth (The Home Depot)); UPX0032 at -147–48 (depicting app ads on google.com and Google Play).

101. Search Ads are typically sold on a cost-per-click (CPC) basis, also referred to as a pay-per-click (PPC) basis, meaning advertisers only pay when a user clicks on a Search Ad. Tr. 1176:6–9, 1177:12–20, 1178:21–22, 1184:1–6, 1195:6–16, 1335:9–13 (Dischler (Google)); Tr. 5430:15–18 (Jerath (Pls. Expert)); UPX0001 at -543; Tr. 3991:18–3992:21 (Juda (Google)); Tr. 401:4–6 (Varian (Google)); Tr. 5119:13–5120:2 (Booth (The Home Depot)); UPX0032 at -144–49, -157, -164–65, -167–70 (noting when Google recognizes Search Ads revenue for different Search Ads). Merely showing a Search Ad to a user is free to the advertiser. Tr. 4016:6–13 (Juda (Google)); Tr. 1177:12–20, 1178:3–1179:7, 1179:10–12, 1184:1–6 (Dischler (Google)) (agreeing “[i]mpressions in paid search are free.”); UPX0001 at -538–39; UPX0842 at -001.

b) Text Advertising

102. General search text ads (Text Ads) are a particular type of Search Ad that are shown in response to a consumer’s real-time query entered on a GSE, matched by keywords selected by the advertiser. Tr. 1185:16–19 (Dischler (Google)); Tr. 3809:13–23 (Lowcock (IPG)); UPX0032 at -144 (illustrating a Text Ad).

103. Text Ads appear similar to organic results and consist primarily of text. Tr. 5379:21–5380:8 (Jerath (Pls. Expert)) (identifying Text Ads as contained in the red rectangle in UPXD103 at 4); Tr. 1179:22–1180:2 (Dischler (Google)) (referring to UPX0012 at .005); Tr. 3810:9–23 (Lowcock (IPG)) (referring to UPX0012 at .005); Tr. 3994:10–22 (Juda (Google)) (discussing UPX0032 at -144); Des. Tr. 21:17–22:1 (Miller (Google) Dep.) (Text Ads are “essentially our traditional search ad business which is primarily text-based. It looks a lot like an organic search result . . . .”).

104. Text Ads typically contain a URL, one to three headlines, and one or two descriptive lines of text. Tr. 1179:22–1180:2 (Dischler (Google)); UPX0012 at .005; UPX0032 at -144; Tr. 3995:7–10 (Juda (Google)); Des. Tr. 60:8–16 (Silverman (Google) Dep.) (Text Ad is “an ad format that shows on Google Search” with “a simplified URL for the partner, a clickable headline, and a line of body text.”).

Generally, no more than seven Text Ads are shown per Google SERP (up to four at top and up to three at bottom). UPX0010 at -052–53; Des. Tr. 41:6– 9, 41:12–43:4 (Fox (Google) Dep.) (detailing the number of Text Ads typically shown on the SERP and changes to these numbers over time); UPX0001 at -533 (There are generally 2–3 Text Ads above the organic results.).

105. Text Ads differ from organic results in at least two key ways. First, with Text Ads, the advertiser determines the ad’s message, whereas with organic results, the GSE determines the result. Tr. 3810:9–23 (Lowcock (IPG)) (referring to UPX0012 at .005); UPX0452 at .006 (Dr. Varian writing, “[G]oogle chooses the creative for organic ads, but the advertiser chooses the creative for the ad.”); PSX00092 at -183, -187–89 (Text Ads “provide[] added control [and] flexibility across auctions,” including to “[c]ustomize your message to your customer’s needs.” (emphasis omitted)); Tr. 5440:6–19. (Jerath (Pls. Expert)).

Second, with Text Ads, advertisers have more control over whether they appear on the SERP via their keyword bidding and campaign selections, and advertisers can change, add, or discontinue Text Ads instantly. UPX0926 at -684–88, -691–98 (IPG training materials explaining different ways to refine when Text Ads are shown); Tr. 5440:6–19 (Jerath (Pls. Expert)); UPX0425 at -281–82.

106. By contrast, advertisers have very limited control over whether or where they appear in the organic results; efforts to improve position among organic listings—referred to search engine optimization, or SEO—can take weeks or months to yield results, if improvement occurs at all. Tr. 5440:2–5441:18 (Jerath (Pls. Expert)); UPX0441 at -802 (internal JPMorgan document: “Paid search copy can be changed and tested much faster than it can be on the organic side, helping us learn quickly what language works so that we can use to better inform other marketing messages.” (emphasis omitted)); UPX0450 at .011 (IPG client presentation comparing paid search and organic search); UPX1131 at -375 (Apple recognizes the importance of paid search because it can be “launched within minutes after Tim [Cook (CEO)] walks off the stage” at a product launch, but with SEO, it “can take a few hours or possibly days for SEO rankings to appear.”); UPX8057 (Google guide on how SEO works).

107. In fact, Google views Text Ads and SEO as complements and recommends that advertisers use them together. PSX00092 at -181, -186, -191 (Google presentation (“SEO + SEM = Better Together”) advising that Text Ads “compliment[] SEO” and that “[t]he combination of paid and organic search creates synergies and the highest share of clicks.”); UPX0425 at -271–79 (describing benefits of an SEO-and-paid-search strategy).

c) Display Advertising, Including Advertising On Social Media

108. Display advertising (or display ads) are pictorial-based ads, such as banner ads or video ads, shown on websites across the Internet. Tr. 1193:2–18, 1195:22–25, 1347:3–7 (Dischler (Google)) (“Banner ad” is (1) a pictorial ad that is a type of display ad and (2) run across websites on the Internet.); Tr. 5378:15–17 (Jerath (Pls. Expert)) (“[D]isplay ads are basically banner ads and also video ads on websites and apps on desktop and mobile phones.”); Tr. 3818:12–3819:5 (Lowcock (IPG)).

109. Ads on social media sites such as Facebook, Pinterest, or TikTok are considered a form of display ad. Tr. 5392:3–5393:9 (Jerath (Pls. Expert)) (“[T]he vast majority of advertising on social media is display.”); Tr. 457:19–22 (Varian (Google)) (Facebook delivers display ads.); Tr. 3838:9–3839:3 (Lowcock (IPG)) (“Social advertising can occur across a number of social platforms, so Snapchat, Pinterest, Facebook, Instagram.”); id. 3839:23–3840:2 (Social media ads are “display-type” ads.); id. 3927:25–3928:14 (TikTok is a social media platform.); Des. Tr. 260:21–261:2 (van der Kooi (Microsoft) Dep.) (“Social ads, in my view, are just a form of targeted display ads. There is a category that is created around it mostly by Facebook, but, really, that’s all it is. It’s a targeted display ad.”). And display ads on social media are sometimes referred to as social ads. Des. Tr. 260:21–261:2 (van der Kooi (Microsoft) Dep.).

110. Indeed, although Search Ads can appear on Facebook, they compose a very small percentage of Facebook’s ads. Tr. 5392:3–5393:9 (Jerath (Pls. Expert)) (Search Ads on Facebook is “very, very small. It’s not really a big phenomenon at all.”); Tr. 8772:13–16 (Israel (Def. Expert)) (Search Ads are “a very small percentage on Facebook.”); Des. Tr. 154:22–23, 155:1–3, 155:5–10, 155:15–156:6, 156:20–159:8 (Levy (Meta) Dep.) (Search Ads accounted for [redacted]% of Facebook ad revenue in April 2021 and [redacted]% in October 2020 (discussing UPX1019 (dep. exs. 17 & 18) and UPX2116 (dep. ex. 16))); UPX1019 at -524 (Facebook “Weekly Ads XFN Report”); UPX2116 at -154 (Facebook “Ads Weekly” revenues); UPX2113 at -789 (listing over 20 places an ad can appear on Facebook, Instagram, and other Meta properties with “Search” far down the list and limited to Facebook).

111. Google also has social ads, called Demand Gen or Discovery Ads or Campaigns.[3] Demand Gen ads appear on Google’s “feed” products, are not displayed in response to a query, and are modeled on the format of ads displayed on Facebook and Instagram. Tr. 1196:15– 1197:5, 1347:11–17 (Dischler (Google)); UPX0033 at -117 (“Discovery ads . . . allow[] advertisers to extend the reach of their social ads to 2.9B+ Google users.”).

Indeed, Google launched discovery ads to compete with social ads, because “[d]iscovery ads provide a familiar pitch to social buyers.” UPX0033 at -145 (2020 presentation) (discussed at Tr. 4646:9–4647:7, excerpted in UPXD102 at 32, and shown in Figure 3, infra ¶ 427); UPX0033 at -130 (“Social buyers want inventory like social feeds.”).

112. Another type of display ad—retargeted display ads—is targeted at consumers who have recently visited an advertiser’s website but who have not taken a desired action (e.g., made a purchase). Tr. 5445:12–5448:9 (Jerath (Pls. Expert)) (discussing UPXD103 at 20). As Dr. Varian acknowledged, a retargeted display ad can only be used after the consumer has visited a merchant’s website. Tr. 455:25–456:5 (Varian (Google)); UPX0414 at -697; UPX0026 at -764; Tr. 5445:24–5447:13 (Jerath (Pls. Expert)) (discussing UPXD103 at 20).

Moreover, “[m]ost of the value of retargeted ads occurs in the first hour or so after the user visits the advertiser’s web page.” UPX0026 at -765; Tr. 456:18–457:17 (Varian (Google)) (acknowledging authorship of UPX0026 and adopting the observation as true); id. 456:6–17 (agreeing the “value of retargeted ads fades over time”).

113. Display ads are typically priced per impression, i.e., when the ad is shown. Tr. 5430:19–5431:5 (Jerath (Pls. Expert)); Tr. 5119:19–5120:2, 5120:15–21 (Booth (The Home Depot)) (“Display is more of a CPM, or a cost per thousand impressions, which is buying a volume of times that an ad is shown”); Tr. 3821:6–20 (Lowcock (IPG)). Advertisers are charged regardless of whether anyone clicks on their ad. Tr. 1195:6–16 (Dischler (Google)) (“[With CPM,] you would pay a certain rate for 1,000 impressions of the ad, 1,000 people who may or may not be seeing the ad but it’s being rendered on the page. In contrast, if you’re on a cost-perclick basis, that means that you have to actually click on the ad, interact with the ad in order to pay for it.”).

2. Targeting Digital Ads

114. Targeting refers to an advertiser’s effort to put their ad in front of the people with whom the advertised product or service would most resonate. Tr. 5389:3–8 (Jerath (Pls. Expert)). Depending on format, advertisers can target digital ads using (1) declared intent in real time via a query or (2) inferred intent from signals. Id. 5389:3–5390:20; UPX0025 at -009 (“Major difference between Search vs Display & YouTube” is “intent” (search) and “audience” (display and video).).

a) Declared Real Time Intent

115. Declared intent in real time refers to when a consumer declares what they want, i.e., their intent. Tr. 5390:6–15 (Jerath (Pls. Expert)). For example, when a consumer types a query into a search engine, they are expressly telling the search engine that, at that exact moment in time, they are interested in the query’s subject matter. Id. 5390:6–15 (likening it to a consumer visiting a store and telling a salesperson what they want at that moment (referencing his testimony at Tr. 5386:6–5387:25)). As the query is a statement of intent, no inferences are needed. Id. 5390:16–20.

116. A consumer query entered on a search engine is a powerful signal of the consumer’s intent and interest. Tr. 4854:5–13 (Lim (JPMorgan)) (Search is one of the strongest intent signals because it responds directly to a query, which is “unique to paid search.”); Tr. 9234:1–8, 9284:6–9 (Holden (Google)) (Users “express[] intent through a query,” and “[s]earch is a powerful intent signal that can be utilized on many different sites.”); Tr. 404:17– 20; 405:18–406:7 (Varian (Google)); UPX0441 at -802 (Search queries are “some of the strongest intent signals made available,” and “search can capture a consumers[’] intent to transact . . . at its highest point.”); UPX0910 at -753 (“The vast majority of our [Google’s] profits come from search ads, because the signal is [so] strong.”); Des. Tr. 56:11–57:14 (Utter (Microsoft) Dep.) (“In a search engine, you are describing your intent with a very firm signal, a query.”); Des. Tr. 137:2–8, 139:2–23 (Levy (Meta) Dep.) (Facebook analysis (UPX1020 at -504 (dep. ex. 13)) in 2020 assumes prices for intent Search Ads (“US Intent Search CPM”) are far higher than for inferred intent ads (“US FB Inferred Intent CPM”).); Tr. 5390:16–20 (Jerath (Pls. Expert)).

117. As Dr. Varian acknowledged, a search query is a very strong predictor of the type of ad content that would be most relevant to the user. Tr. 404:17–406:7 (Varian (Google)) (agreeing that query for “cheap blue blender” is a signal the consumer is interested in cheap blue blenders). This is especially valuable because the advertiser’s goal is to connect with consumers who have expressed interest in the advertiser’s product or service. Tr. 9236:8–15 (Holden (Google)) (“The [advertiser’s] goal is to connect . . . with a user who has expressed interest in their product or service.”).

b) Inferred Intent

118. When using inferred intent, advertisers look for signals in consumer data from which the advertiser could determine the subject matters in which a consumer may be interested. Tr. 5389:9–5390:5 (Jerath (Pls. Expert)) (discussing UPXD103 at 7); Des. Tr. 284:20–285:6 (Alberts (Dentsu) Dep.) (Implied intent “needs to be inferred based on what we know about a consumer and they may not expressly share what their intent is.”).

This is sometimes called “audience targeting.” UPX1005 at -188 (Google document on “Search Ad Themes”); UPX8016 (Google Ads Help: “About audience targeting”); UPX0913 at -021 (Google “Online Ads” presentation by Dr. Varian). Such signals in the data (or data signals) include expected audience composition (e.g., demographics), behavioral profiles based on past or recent online behavior, and context. Tr. 5389:9–5390:5 (Jerath (Pls. Expert)) (discussing UPXD103 at 7); Tr. 1418:17– 1420:3 (Dischler (Google)) (User’s website visits and video views are signals of the user’s intent.); UPX0926 at -684–87 (IPG training materials on targeting Search Ads); UPX8019 (Google Ads Help: “About demographic targeting”); UPX8027 (Google Ads Help: “About targeting geographic locations”).

119. Although signals from which intent is inferred can be useful, they also can be low information and limited in accuracy. Tr. 5389:9–5390:5 (Jerath (Pls. Expert)). Thus, if a person is reading an article about an earthquake in Haiti, that does not mean they are interested in vacationing there. UPX0026 at -772 (Dr. Varian’s “Online advertising primer”: “[I]t is almost impossible to contextually target news. What do you show next to a story about a hurricane in Haiti? Travel ads for Haiti? Probably not.”).

Similarly, the signal may be old, making it unclear whether the consumer is still interested in what they saw previously. Tr. 5389:9–5390:5 (Jerath (Pls. Expert)); Tr. 426:7–13 (Varian (Google)) (agreeing the farther away in time “from a search or an indication of interest in a product, the less value that signal has”). For example, a consumer who searched for a tennis racquet three days ago may no longer be interested in tennis or may have already purchased a racquet. Tr. 424:19–426:13 (Varian (Google)); UPX0910 at -753.

c) Traditional, Display, And Social Ad Channels Are Based On Inferred Intent And Are Most Suited For Generating Demand

120. Traditional advertising is targeted based on inferred intent from signals, such as expected audience composition or demographics of who may be interested in the advertiser’s product or service. Tr. 5389:9–5390:5, 5391:10–23 (Jerath (Pls. Expert)) (discussing UPXD103 at 7–8); Tr. 3826:11–3827:11 (Lowcock (IPG)) (TV, radio, and out-of-home (e.g., billboard) ads build awareness and interest among consumers, resulting in searches.).

121. Display ads are also targeted based on inferred intent from signals, often behavioral profiles based on online behavior. Tr. 5389:9–5390:5, 5391:10–23 (Jerath (Pls. Expert)) (discussing UPXD103 at 7–8). Indeed, as Dr. Varian has explained, display ads lack the strong intent signal of a query and therefore must be targeted based upon other information, such as context or demographics. Tr. 427:21–428:24 (Varian (Google)) (discussing UPX0910 at -753); UPX0910 at -753 (Dr. Varian writing, “The vast majority of our profits come from search ads because the signal from the query is [so] strong.” And lacking strong signals, publishers (i.e., sellers of display ads) “need to target based on context or demographics.”); UPX0913 at -021 (Dr. Varian presentation listing ways to target display ads).

122. Like other display ads, social ads are targeted based on inferred intent from signals, such as groups in which a consumer has shown interest. Tr. 5238:3–5239:6 (Dijk (Booking.com)) (Booking.com uses Facebook ads to build awareness and consideration by targeting visitors of travel-focused groups.); Tr. 7386:10–7387:13 (Raghavan (Google)) (Social media targets ads using “latent intent” signals.); Des. Tr. 118:7–119:10 (James (Amazon) Dep.) (“[O]n Facebook the targeting is done in a different way.

That is, it is more creating an ad which is really more focused around the graphic nature of the products and that it is determined by Facebook as to whether or not that ad is relevant for the user. But it is not driven by a query, perse.”). Thus, on TikTok, users do not typically enter a search to view content but rather scroll through an algorithmic video feed created based upon a user’s engagement with previous videos. Tr. 7419:9–7420:15 (Raghavan (Google)).

123. Traditional, display, and social ads are most suited and effective for goals associated with generating demand, e.g., increasing awareness and interest, which fall at the top and middle of the consumer purchase funnel. Tr. 5391:10–23, 5443:1–5444:15 (Jerath (Pls. Expert)) (discussing UPXD103 at 8, 19).

124. Display ads “are essentially an advertiser putting a message out there when a consumer isn’t necessarily even looking for something.” Tr. 5123:1–5424:1 (Booth (The Home Depot)) (specifically mentioning banner and social ads); Des. Tr. 149:18–19, 149:21–22 (Ramalingam (Yahoo) Dep.) (“[O]ne of the primary objectives” of display ads is to build brand awareness.); Des. Tr. 252:25–253:10 (van der Kooi (Microsoft) Dep.) (Display ad is generally used for “awareness-based advertising.”). Thus, ad agency IPG views display ads as most effective in driving awareness. Tr. 3816:7–11, 3819:6–9 (Lowcock (IPG)).

125. Advertiser’s views are consistent with IPG’s view. The Home Depot uses display and social ads to nurture the consumer journey by presenting the consumer with multiple product and service options, with the goal of leading the consumer down the transaction path to where paid search focuses. Tr. 5122:1–20 (Booth (The Home Depot)).

“If General Motors launches a new vehicle and they want to put that new vehicle in front of as many consumers as they can, they buy lots of display ads that are out on the Internet.” Des. Tr. 254:1–4, 254:9–255:9 (van der Kooi (Microsoft) Dep.).

126. Similarly, Booking.com advertises on Facebook, Instagram, and TikTok to drive awareness and consideration. Tr. 5241:17–5243:17, 5279:21–5280:9 (Dijk (Booking.com)). And others in the advertising industry do the same. Tr. 5120:22–5121:25, 5168:4–25 (Booth (The Home Depot)) (The display and social teams focus “more on the upper funnel, where they’re trying to inspire, they’re trying to bring awareness, they’re trying to kind of nurture that consumer path, to the point where they’re eventually at a point where they want to make a transaction.”); Tr. 6587:4–17 (Vallez (Skai)) (Skai views social ads being “more upper funnel . . . where we’re trying to influence and engage [with consumers] when they’re not necessarily in the context of purchasing.”); UPX0445 at -507 (Facebook executive: “Our strength (core) is the top of the funnel.”); UPX2114 at -761 (Facebook VP of Ads Engineering[4] in 2021: “[P]eople don’t come to core FB products with intent (unlike search), from the beginning, we knew we were about demand generation, not demand fulfillment . . . .”); UPX0916 at -765 (Facebook Group Lead in 2016 explaining why travel advertisers spend more on Google than Facebook: “We don’t have the lower funnel intent that Google delivers via search.”).

127. Although traditional, display, or social ads can reach consumers in other parts of the funnel, they are not as well suited and effective for achieving those goals compared to top funnel goals of awareness and consideration. Tr. 5391:24–5392:2 (Jerath (Pls. Expert)).

For example, IPG views display ads as “not effective at all stages of the funnel” and “not effective[]” at leading to purchases. Tr. 3929:18–3930:17 (Lowcock (IPG)); id. 3839:23–3840:2, 3933:11– 13, 3980:5–3981:6. Similarly, none of JPMorgan’s paid social ads in the last year was for acquisition. Tr. 4905:6–4906:25, 4915:25–4916:2 (Lim (JPMorgan)).

Other industry participants have consistent views. Tr. 5241:17–5243:17 (Dijk (Booking.com)) (Booking.com advertises on Facebook, Instagram, and TikTok to drive awareness and consideration, but not to reach highintent users.); Tr. 6513:1–24 (Hurst (Expedia)) (Goal of social ads is engagement because of “intent and where someone is in a purchase decision . . . . [W]e would rarely know in a social channel if someone is trying to book travel right now.”).

d) Advertisers Attach A Unique Value To Text Ads And Other Search Ads, Which Are Most Suited And Effective For Harvesting Demand

128. Advertisers use different types of advertising, or ad channels, for different purposes. Tr. 3814:19–3815:13 (Lowcock (IPG)) (explaining IPG training materials, UPX0926 at -684, “Orienting these campaigns with the customer journey is critical so that you can align all assets housed within the campaign to a common and consistent goal.”); Tr. 5390:21–5391:9 (Jerath (Pls. Expert)) (discussing UPXD103 at 7, “advertisers really want to figure out and understand[,] . . . for meeting this particular goal of the campaign, which channel is most suited and most effective.”).

129. Text Ads and other Search Ads have a unique value to advertisers, allowing them to target consumers’ declared intent in real time and harvest demand for their products and services. Tr. 5443:1–5444:15 (Jerath (Pls. Expert)) (discussing UPXD103 at 19). Only Text Ads and other Search Ads are targeted based on consumer’s declared intent, in real time. Tr. 5391:10–23 (Jerath (Pls. Expert)) (discussing UPXD103 at 8); Tr. 5220:9–22 (Booth (The Home Depot)) (“[S]earch ads are unique in the sense that somebody is going and initiating the action, going to Google, going to Bing, asking for solutions to -- asking for a solution.”); Tr. 9076:6–17, 9077:5–14 (Fitzpatrick (Google)) (agreeing with Google’s “search explainer” Danny Sullivan’s statement, written in UPX1044 at -718, that “[t]he query itself—not any data about the user—is by far the most powerful signal for which results are most relevant and useful.”).

130. Google recognizes the unique value of search and the differing purposes of other ad channels, and Google markets its own ad products accordingly. It identifies its search campaigns as “Google’s best performing campaign, the first place you should spend and maximize budgets.” UPX0451 at -956. Google’s other products are similarly categorized: Performance Max Campaigns are a bundled product second in importance to Search Campaigns, Discovery Campaigns are “the best way to spend social budgets with Google,” Video Action Campaigns are “the best way to spend performance video only budgets,” and Google Display Ads Campaigns are “the best way to spend display only budgets with Google.” Id.

131. As Tracy Lim, Managing Director and Chief Media Officer for JPMorgan Chase, explained, “[B]ecause we are not intercepting you whilst you’re enjoying your TV show or whatever it is that you may be doing on another channel, you have come to Google or Bing as a consumer and typed in your query. We’re responding to that query, which is unique to paid search.” Tr. 4854:4–13 (Lim (JPMorgan)); Other industry participants have similar views. Tr. 5123:1–23 (Booth (The Home Depot)) (With “pull ads,” i.e., paid search, the consumer “is actively looking for something,” and The Home Depot has “the opportunity to be able to respond to that query” and “bring[] people in who are already in market.”); Des. Tr. 254:1–4, 254:9– 255:8 (van der Kooi (Microsoft) Dep.) (“Pull advertising is . . . a consumer pulls the ad by typing into the search engine ‘electric vehicle from General Motors.’

When they type that in and see a search ad for an electric -- for a Volt -- the Chevy Volt, then they pull the ad down by way of their query.”); UPX8033 at .001 (Google Ads Help: “Create a Search campaign,” listing benefits of Search Ads and including the ability to “[t]arget people actively searching for your specific products and services.”); UPX1014 at -079 (Search Ads has the “ability to ‘pull’ users into a brand’s message.”).

132. Marketing platform Skai similarly expressed that “[t]here’s no paid media channel that better captures the intent of users” than Search Ads. PSX01200 at 1 (Skai webpage noting the importance of Search Ads); Tr. 6585:25–6586:17 (Vallez (Skai)) (agreeing with statement in PSX01200 at 1).

133. Other industry participants likewise believe that Text Ads and other Search Ads are uniquely able to target consumer’s real-time declared intent. UPX0450 at .006 (IPG highlighted for a client that “search reaches consumers at the moment of intent.”); Des. Tr. 140:14–17, 140:19–141:14, 141:16–17 (Ramalingam (Yahoo) Dep.)

(Search Ads’ unique value proposition is that they have “higher purchase intent compared to any other form of advertising.”); Des. Tr. 284:7–285:22 (Utter (Microsoft) Dep.)

(The query has “strong and very unique intense signal” accessible by Search Ads.); Tr. 5241:2–11 (Dijk (Booking.com)) (agreeing with statements in ¶ 29 of Plaintiff’s Amended Complaint (shown in DXD-03 at .006) that Search Ads enable advertisers to target potential customers at the exact moment users express interest in a topic.); UPX0026 at -764 (Dr. Varian’s “Online advertising primer”: “Search ads are an effective form of advertising since queries are a strong signal of user interest and intent and the ads appear immediately after the query is entered.”).

134. Given their ability to target real-time declared intent, Text Ads and other Search Ads are most suited and effective for harvesting demand. Tr. 5391:10–23 (Jerath (Pls. Expert)) (discussing UPXD103 at 8).

For example, Ms. Lim described Text Ads as “the only channel that [JPMorgan Chase] consider[s] . . . [a]s exclusively [a] demand channel” and further confirmed in responding to the Court that “within paid advertising, [paid search] is the biggest acquisition driving channel.” Tr. 4841:2–12, 4853:1–6, 4854:22–4855:6 (Lim (JPMorgan)); UPX0441 at -801–02 (JPMorgan “paid search deep dive doc”: “Search is a consumer demand driven media channel, through search queries it has the ability to answer people’s questions with expansive coverage on 100Ks+ keywords. Search can drive acquisition based on some of the strongest intent signals made available.” (emphasis omitted)).

Similarly, Global Chief Media Officer Joshua Lowcock, at Interpublic Group’s Universal McCann, explained that Text Ads are most effective at driving conversions. Tr. 3816:23–3817:1, 3826:11–3827:11, 3835:25–3836:10 (Lowcock (IPG)); id. 3836:15–3837:4 (Search Ads are not an “effective method to driving awareness,” and they are highly cost ineffective for branding.); UPX0450 at .006–07 (IPG client presentation: “[P]aid search is operating in the lowest part of the funnel, capturing any demand driven from upper funnel activity . . . .” (emphasis omitted)).

135. Booking.com similarly purchases Search Ads to get “high-intent customers” who “have expressed a very clear interest in booking a hotel.” Tr. 5236:19–5237:1 (Dijk (Booking.com)).

Other industry participants likewise believe Search Ads are the best method to target user intent. Tr. 6586:18–6587:3 (Vallez (Skai)) (“[S]earch is more often than not the last step, one of the last steps in that [purchase] journey. . . . [S]earch again is more often than not that channel where they make their [purchase] decision.”); Tr. 5120:22–5121:25 (Booth (The Home Depot)); Des. Tr. 45:4–46:16 (Alberts (Dentsu) Dep.) (“Search can be great way to capture demand that’s at the bottom of the funnel.”); Des. Tr. 108:2–109:9 (Raymond (Kohl’s) Dep.) (“Oftentimes search is one of the . . . last places a customer will go.”); Des. Tr. 121:18– 122:2 (James (Amazon) Dep.) (“With respect to the search ads program, Text Ads and Shopping ads, intent is an important signal in terms of what that would lead to from a conversion perspective.

A user who is searching for the Dyson DC65 vacuum is probably very interested in purchasing that specific model of vacuum. So, yes, it is an important signal in terms of the conversion probability.”).

136. The different ways in which ads are priced reflect how different ad channels align with different goals. Indeed, display ads are typically priced per impression, corresponding to the goal of generating demand, but Text Ads and other Search Ads are priced on a click-basis, corresponding to the lower-funnel goal of harvesting demand. Tr. 5430:15–5431:5 (Jerath (Pls. Expert)).

137. Average ad click rates similarly reflect how different ad channels align with different goals. On average, an advertiser can expect consumers to click on their display ad in the range of only 1–10 times per 10,000 impressions. Tr. 1194:25–1195:13 (Dischler (Google)) (Display ads are sold on a cost per mille (CPM) basis, i.e., a cost per thousand impressions.); Tr. 5431:6–14 (Jerath (Pls. Expert)). By contrast, on average, an advertiser can expect consumers to click on their Text Ad in the range of 200–400 times per 10,000 impressions—which is as much as two orders of magnitude greater than for display ads. Id. 5431:15–5432:2.

3. Google’s Search Ads Products And Their Development

138. Google offers up to seven ad “positions” for purchase on each of its mobile and desktop SERPs—four in the “top slot” (i.e., above the organic results) and three in the “bottom slot” (i.e., below the organic results). UPX0010 at -051–53. Each of the seven positions can house a single Text Ad, or a “carousel” of shopping ads can appear in one of the top slots.

UPX0025 at -028 (“Shopping ads appear on top of organic search listings . . . .”); UPX0012 at .003–04 (illustrating desktop and mobile SERP and showing shopping ads in top slot on both); UPX6058 at -002 (Google Ads Help: “About ad position and Ad Rank”: “Ad position is the order of your ad in the auction results compared to other ads.”). Desktop SERPs also display shopping ads—but not Text Ads—along the right-hand side of the SERP. UPX0464 at -157 (illustrating desktop SERP with shopping ads); Tr. 4617:1–17 (Whinston (Pls. Expert)) (describing depictions of SERPs from google.com and amazon.com in UPXD102 at 10, depicting Google SERP with shopping ads on the lower right side).

139. Not every search query on Google returns a Search Ad; rather, Google returns ads only for queries with “commercial intent.” Commercial intent is something of a truism: a query has “commercial intent” if an advertiser is willing to advertise on it. Tr. 1171:3–21 (Dischler (Google)); id. 1171:23–1172:1; Tr. 398:13–15 (Varian (Google)) (“A commercial query is one that’s related to business or commercial matters.”); Tr. 3686:22–3687:8 (Ramaswamy (Neeva)) (“[C]ommercial intent covers queries that people are willing to pay money for.”).

a) Auctions

140. For each search query, Google auctions off available ad space on its SERP. Google identifies ads targeting the query from millions of candidates; the targeting method varies by ad format. If any eligible ads exist, Google selects, ranks, and prices them using a variant of a second-price auction. Tr. 4011:16–23, 4015:17–22, 4240:21–4241:22 (Juda (Google)); Tr. 1199:25–1200:3 (Dischler (Google)); DX0153 at -092.

Google separately auctions off each ad position on the page, first auctioning off the top position, then retrieving a new set of candidates and running a new auction for the second position, and so on. DX0153 at -092; UPX0010 at -064–66 (Google regulatory submission explaining the auction).

Moreover, Google maintains wholly separate auction stacks for Text Ads and other Search Ads, meaning Text Ads are not auctioned against PLAs, nor vice versa. Tr. 4018:24–4019:3, 4019:10–18 (Juda (Google)); Tr. 1197:9–13 (Dischler (Google)); Tr. 3812:9–15 (Lowcock (IPG)); UPX6032 at -654–55 (Google admission (Resp. 11) that it has different auctions for shopping ads and for Text Ads). Thus, a single query can trigger multiple Google-run auctions.

141. In a classic second price auction, the highest bidder wins but pays a price equal to or slightly above the first runner-up’s bid, making the runner up’s bid critical to the ultimate price paid by the winner. Tr. 1200:4–25 (Dischler (Google)). Google calls its variant a “Generalized Second Price Auction.” Tr. 4015:17–22 (Juda (Google)); DX0153 at -092.

142. Instead of ranking and pricing by bid, for each ad Google calculates an LTV (long-term value) score—referred to externally as Ad Rank—using the advertiser’s bid and several Google-calculated assessments of the quality of the ad. DX0153 at -092; UPX8030 at .001 (Google Ads Help on Ad Rank); UPX0010 at -054–57 (overview of LTV and Ad Rank). Google then ranks and prices the ads in each position by LTV/Ad Rank. DX0153 at -092; UPX8030 at .001 (Google Ads Help on Ad Rank); UPX0010 at -054–57 (explains iterative pricing and ranking process).

143. Google’s design and control of the auction, coupled with the auction’s reliance on internal calculations, gives Google the ability to affect the auction’s outcome, including the price ultimately paid by advertisers. Indeed, and as conceded by Dr. Varian, a monopolist selling through an auction process has multiple levers through which it can affect the outcome. Tr. 465:9–12 (Varian (Google)). Google uses these levers to conduct “intentional pricing,” which it defines as changing the auction’s design to change prices. UPX0509 at -869; Tr. 4102:18– 4103:2 (Juda (Google)) (Google can directly change how the auction works, impacting pricing.). Infra ¶¶ 629–637 (§ V.C.5.a).

b) Launches

144. Google internally refers to changes to its Search Ads business—i.e., changes in the appearance of the SERP or Search Ads themselves (including Text Ads) or changes to the various Search Ads auctions—as “launches.” Des. Tr. 15:6–11 (Miller (Google) Dep.); Tr. 1205:19–25 (Dischler (Google)). For each change, Google creates and maintains documents, known as “Launch Docs,” reflecting the launch details. Des. Tr. 159:4–20 (Jain (Google) Dep.) (discussing UPX0746 (dep. ex. 9), a launch doc for the Holy Load launch).

145. Google’s ad launch process uses live Google traffic to assess a launch’s impact on Google’s revenue and user metrics. Des. Tr. 15:12–13, 15:15–24, 16:2–14, 16:20–17:6 (Miller (Google) Dep.). This includes an escalating series of experiments, which start with simulations or A/B tests conducted on very low levels of live traffic. Id.; Tr. 4271:22-4272:15 (Juda (Google)); Tr. 2315:15–2316:9 (Giannandrea (Apple)) (Google did not "typically” make a change to its algorithm without a live A/B experiment (testifying about his time at Google)); Des. Tr. 92:3-12 (Jain (Google) Dep.).

Assuming a positive outcome, the live experiments will be expanded to include larger levels of traffic, eventually reaching 10% of traffic, at which point more senior “launch approvers" need to sign off on the experiment. Tr. 4271:22-4272:15 (Juda (Google)).

Assuming continuing good performance, the experiments will be expanded further, including at least a step involving 50% of traffic, before finally going to all traffic. Id. Although Google also attempts to conduct post-launch analysis of effects, it believes its pre-launch process more accurately measures the effect of a change because of “noise in the system” after a full launch. Des. Tr. 15:12-13, 15:15-24 (Miller (Google) Dep.).

---

[3] Demand Gen ads were formerly known as Discovery Ads, but Google recently rebranded them. Tr. 7537:20–7538:8, 7543:17–7544:3 (Raghavan (Google)) (Discovery ads were renamed “Demand Gen” ads and are aimed at buyers of social ads.).

[4] Des. Tr. 113:15–21 (Levy (Meta) Dep.) (identifying Ning Li’s position).