![Search icon](https://hackernoon.imgix.net/search-new.png?w=19&h=19)
United States of America v. Google LLC., Court Filing, retrieved on April 30, 2024, is part of HackerNoon’s Legal PDF Series. You can jump to any part of this filing here. This part is 30 of 37.
1160. Due to reduced competition in the United States, Google has not implemented Search Ads launches or improvements that would have benefited advertisers. Tr. 5862:18–5863:13 (Whinston (Pls. Expert)) (“[W]hen Google doesn’t face any competition, it sees itself losing revenue by doing these quality improvements for advertisers, so it doesn’t do them.”).
1161. Lack of competition reduces Google’s incentives to respond to advertiser preferences, because advertisers lack a viable alternative to Google. Tr. 5859:20–5860:3 (Whinston (Pls. Expert)); id. 5862:18–5863:13 (“If Google face[d] competition and it doesn’t do the quality improvements, it will lose advertisers to rivals. The whole calculus changes, and now the rational business decision when it faces competition is to do the ad launch . . . . [Competitive markets] force firms, they push firms [so] that their rational business decisions are to do things that are good for consumers; in this case advertisers.”).
incentives to do things that are good for advertisers, and that could be quality improvements, and it could be price reductions, price reductions or not increasing price as much.”).
1163. Google internally ascribes lower RPMs in Japan to the fact that Google faces competition from Yahoo in Japan. UPX0462 at -844 (one “take away” as to why RPMs are lower in Japan is that “Japan is unique, we have a big competitor unlike US and UK”); Tr. 1550:22–1552:4, 1554:19–1556:21 (Roszak (Google)) (discussing UPX0462).
1164. Unlike the United States, advertisers in Japan benefit from lower pricing because they can use Google’s competitor (Yahoo) as an alternative. UPX0462 at -844 (“I’d highlight the competitive situation in JP [Japan]. JP is unique among our major countries in a sense that we have a player who competes against us head-to-head, Y!J [Yahoo Japan] (even though we surpassed Y!J in search revenue ~2 years ago). Advertisers split their search budget to Y!J and Google, which makes the auction pressure on Google less.
So I don’t think it’s much apple-toapple to compare RPM in JP against RPM in US/UK where we don’t have competitors like Y!J, and I’ve rather compared our RPM against Y!J’s (even though we need to make lots of assumptions to estimate it)”). Reduced auction pressure in Japan reduces CPCs. Supra ¶ 622.
a) Google Offered Advertiser-Friendly Incentives In Japan When Faced With Competition From Yahoo
1165. Google offered advertiser-friendly incentive programs when responding to unique competition in Japan. In approximately 2010, Google faced, from Yahoo Japan, a meaningful competitive threat for Search Ads. Google, accordingly, implemented an “incentive program” called “JIP [Japan Incentive Program] Search,” which effectively operated as a Search Ads discount program aimed at ad agencies. UPX0057 at -846; Tr. 5860:4–5861:6 (Whinston (Pls. Expert)) (presenting UPXD104 at 80 (citing UPX0057 at -846)).
1166. Google implemented JIP Search to “help [its] competitive position against Y! who used to be the largest player in the Japan online ads market and offered an incentive program for agencies.” UPX0057 at -846; Tr. 5860:4–5861:6 (Whinston (Pls. Expert)) (presenting UPXD104 at 80; citing UPX0057 at -846)); UPX0785 at -890 (JIP Search provides “[n]on-standard incentives”; “[r]ationale” is that “[w]e have unique competition against Yahoo in Japan”).
1167. Google began phasing out the JIP Search incentive program after its market share grew in Japan and it had “less reason to have JIP Search from the competition perspective.” UPX0057 at -846. In 2014, Google surpassed Yahoo Japan’s search revenue and, as of 2020, Google exceeded Yahoo Japan by approximately three times. UPX0057 at -846. Accordingly, Google began diminishing the scope of JIP Search and, in 2020, was planning to sunset the product. UPX0057 at -846; Tr. 5860:4–5861:6 (Whinston (Pls. Expert)); UPX0053 at -982 (“our intent is actually to terminate JIP Search in the future”) (emphasis in original).
1168. Google considered expanding the Japan Incentive Program to the United States, but did not do so. Tr. 5860:23–5861:6 (Whinston (Pls. Expert)) (describing failure to expand JIPlike program to the United States); UPX0053 at -982 (discussion on expanding JIP search outside of Japan and stating “[r]egarding search incentives, that was something Scott brought up during the discussion, to potentially temporarily increase/introduce search incentives but Philipp [Schindler, Google’s Sr. VP & Chief Business Officer] didn’t really engage much (pointing to how difficult it usually is to get rid of incentives once we have them), and the discussion moved [on]. I don’t think this topic will get much traction.”).
1169. Over time, Google has steadily reduced the granularity and visibility it provides advertisers into their own ad spend and performance, reducing the quality of Google’s ad products.
a) Google Reduced Advertiser’s Visibility Into Their Own Ad Spend Through Changes To Search Query Reports
1170. Google’s search query reports (SQR) identify the queries matching the advertiser’s ads, the average CPC for those ads, the total amount spent, and other metrics. UPX0526 at -538 (SQR background and excerpt); Tr. 4864:23–4865:16 (Lim (JPMorgan)); Des. Tr. 161:2–11 (Alberts (Dentsu) Dep.) (SQRs show specific queries matching keywords.); Tr. 5468:22–5469:8 (Jerath (Pls. Expert)) (discussing UPXD103 at 35; showing excerpt from Google Ads Help webpage)).
Advertisers have no other means of acquiring this information. UPX0987 at -127 (“Particularly given the rise of close variants over the last couple of years, advertisers have increasingly been reliant on search query reports to find poor matches and prevent them with keyword negatives. Any limitations to how many queries are being reported is an obstacle to such control.”)
Information, such as that in the SQR, is a quality aspect of the product purchased by advertisers. By deciding what data is in the SQRs, Google controls what advertisers know about their ad spend. Tr. 5469:18–5470:6 (Jerath (Pls. Expert)).
1171. SQRs are “widely used by advertisers of all segments.” UPX0526 at -539. Advertisers use SQRs to optimize their ad campaigns, including by (1) analyzing keyword performance, (2) adjusting bid strategies, and (3) identifying poorly performing queries. Tr. 5469:9–17 (Jerath (Pls. Expert)); UPX0926 at -691–93, -700–01 (IPG training materials explaining how to use SQRs); UPX8035 at .001 (Google Ads Help webpage: “Evaluating ad performance on the Search Network,” advising that “[t]racking statistics like clicks and impressions is a great way to start”); UPX0058 at -771 (advertisers use SQRs for “[u]nderstanding [] what queries they are appearing against,” “[m]ining for negative keywords,” [c]reating more tightly themed ad groups and adding new keywords,” and “[c]hecking their agency’s targeting,” among other things); UPX0532 at -567 (Google “Comms Doc” discussing SQR uses).
For example, JPMorgan uses SQRs “to inform how [one] keyword or keyword set is performing relative to another,” “to understand which keywords are driving the most value for the firm,” and “to inform whether or not we need to increase or decrease budgets against certain keywords.” Tr. 4865:17–23, 4866:6–9 (Lim (JPMorgan)).
Other advertisers do the same. Tr. 3846:23–3847:12 (Lowcock (IPG)) (SQRs assist in determining the keywords to bid on and evaluating their effectiveness.); Des. Tr. 161:12–162:6 (Alberts (Dentsu) Dep.) (SQRs are used “to determine if we should add new keywords . . . or if we should omit keywords . . . via negative keyword additions.”).
b) Google Harmed Advertisers By Removing The “One-Click Threshold” From The SQR
1172. In September 2020, Google reduced the information provided in SQRs by including fewer queries and by providing advertisers less data and details about their ad spend. Tr. 5469:18–5470:6 (Jerath (Pls. Expert)); Tr. 3849:14–19 (Lowock (IPG)) (Google has deprecated or limited certain types of information available in SQRs.); Des. Tr. 213:13–214:6 (Alberts (Dentsu) Dep.) (Post-2020 SQRs “not as granular and comprehensive of a list as what we used to have.”).
1173. In particular, Google removed the SQR’s “one-click threshold,” where the SQR included any query resulting in an ad click, even if only clicked once. Tr. 5473:25–5474:22 (Jerath (Pls. Expert)) (discussing UPXD103 at 36; citing UPX0526 at -545, -549); Des. Tr. 173:20–174:3 (Alberts (Dentsu) Dep.). Post-change, Google also omitted any query from the report if it received less than “[redacted], even if the query received ad clicks.” UPX0532 at -566–67. Google did not disclose the new [redacted] threshold to advertisers.
UPX0532 at -568 (Google “Comms Doc” stating Google does not reveal the threshold number for inclusion in SQR); Tr. 5222:2–19 (Booth (The Home Depot)) (unaware of the threshold); Des. Tr. 166:17–25 (Alberts (Dentsu) Dep.) (does not know the threshold and does not recall Google ever sharing this information); UPX0058 at -772 (Google also did not disclose the previous lower threshold of [redacted].).
1174. Google subsequently charged advertisers for ads displayed on certain queries without telling the advertisers what those queries were, what the average cost per click was, or what the total spend for each of those queries was. Tr. 5469:18–5471:12 (Jerath (Pls. Expert)) (“[Advertisers] were buying certain queries but they were not being told what they’re buying, which queries they’re buying”; “[T]his is like if you buy a product in a supermarket but they don’t tell you what you actually bought.”); Tr. 5174:9–25 (Booth (The Home Depot)) (SQR changes led to loss in visibility of CPC data.).
1175. The SQR deprecation ultimately concealed anywhere from 20% to greater than 28% of an advertiser’s paid search spend on Google. Tr. 5471:14–5472:1 (Jerath (Pls. Expert)) (“[A] broader study . . . estimated [the figure] to be about 28 percent of ad spend,” while “in [his] personal experience with certain advertisers, [he saw] numbers even greater than that”); Des. Tr. 168:25–169:3, 169:7–12, 169:21–170:12 (Alberts (Dentsu) Dep.) (SQR change resulted in “measurable impact” in terms of the loss “of our visibility into a percentage of queries that were driving costs and clicks[.]”).
1176. For example, JPMorgan saw an increase from only 5% of keyword performance not being visible at the keyword level, “to roughly 20 percent” not being visible. Tr. 4866:9– 4868:5 (Lim (JPMorgan)). Tinuiti, an advertising agency with a focus on digital marketing, saw “a significant drop in the share of spend attributed to search queries across ad formats and device types from August to September” among dozens of advertising clients spending millions on Google Search Ads. UPX0987 at -124–26.
1177. Google’s SQR changes impede advertisers’ ability to assess and optimize their ad spend and manage their costs. Tr. 5475:2–5476:1 (Jerath (Pls. Expert)) (discussing UPXD103 at 37–39, showing excerpts of UPX0983 at -162–63, UPX0987 at -124–26, and UPX0511 at -611).
Advertisers use data in the SQRs to identify queries on which they do not want to bid or appear and identify “negative keywords” to use to avoid appearing on those queries. Tr. 5469:9– 17, 5472:11–5473:3 (Jerath (Pls. Expert)) (Advertisers used SQRs “to ex post figure out that, well, I don’t want this query, I want to exit this particular auction.”); Tr. 5171:23–5172:15 (Booth (The Home Depot)) (“[SQRs] also help[] us understand, are there certain things that we wouldn’t want to be bidding on . . . .”); UPX0926 at -691 (importance and use of negative keywords); Des. Tr. 131:13–133:6 (Alberts (Dentsu) Dep.) (Visibility reveals keywords that the advertiser wants to suppress.).
1178. If an advertiser does not know that its ad is being matched to certain queries, the advertiser cannot avoid those queries. Tr. 5472:11–5473:3 (Jerath (Pls. Expert)) (“[W]ithout knowing which queries you’re being matched with, you can’t exit these auctions.”); Tr. 5172:20– 5174:20 (Booth (The Home Depot)) (“[L]ess information didn’t allow us to be as thorough in what we would -- would have otherwise done.”).
1179. Advertisers understand how the reduction in SQR visibility harms their ability to identify and use negative keywords. Supra ¶ 1172. Advertising agency Tinuiti observed, “[t]his is a massive decrease in query visibility, making it more difficult for paid search marketers to effectively identify poor-matching queries to weed out via keyword negatives.” UPX0987 at -126. Bank of America similarly noted that Google’s September 2020 change to SQRs would “inhibit advertisers from being able to weed out click and mismatched intent.” UPX0983 at -163.
1180. On average, advertisers’ inability to avoid auctions leads to thicker auctions and higher prices. Tr. 5472:11–5473:3 (Jerath (Pls. Expert)) (“So on average, what this would lead to is advertisers entering more auctions, which leads to thicker auctions, that is more participants in the auction, which would then lead to higher prices.”); supra ¶¶ 622–623.
1181. Other industry participants noted that Google’s changes in SQRs otherwise impeded advertisers’ ability to assess and optimize their ad spend and manage their costs. Tr. 4865:17–23, 4868:7–10 (Lim (JPMorgan)) (Google’s changes “gave my team less information to work with.”); UPX0987 at -126 (According to Tinuiti, the loss of query visibility “also makes it more difficult to identify new query variations driving traffic which might be performing well and should be launched as new keywords.”) Des. Tr. 156:12–22, 157:3–158:2, 158:4–159:12 (James (Amazon) Dep.) (“So when the search query report changes were made, it limits the amount of information that we can have related to the number of unique search queries that occur for a [] given term and the number of keywords that are in our auction that would have triggered as a result of that”; “It impacts optimization insofar as there may be areas where we should be creating text ads to target very specific search queries where we may be able to capture more queries and to do so in a more refined way.”); UPX0511 at -611 ([redacted]).
1182. Even before making the September 2020 changes to SQR reports, Google recognized that the changes would result in substantial data loss to advertisers. UPX0526 at -545 (internal Google presentation warning that “Customer relations: Data loss from SQR will be substantial if 1-click clause is removed”); Tr. 5473:25–5474:22 (Jerath (Pls. Expert)) (discussing UPXD103 at 36). Google further recognized that the “[r]isks of showing less data in SQR[s]” included “[d]isrupt[ing] keyword harvesting algorithms or negative keyword setting workflows” and “[s]maller advertisers los[ing] more data [therefore making] Google Ads [] harder to use.” UPX0058 at -779.
c) Google’s SQR Changes Exacerbate Harm Caused By Changes To Keyword Match Types
1183. Google’s September 2020 changes to its SQRs exacerbated issues, supra ¶¶ 609– 626 (§ V.C.4), related to expanded keyword match types. As described more fully there, beginning in 2012, Google changed its keyword match types to reduce the precision with which advertisers could target queries and denying advertisers the option to opt out of the changes.
Although Google advised advertisers to use negative keywords to deal with the expansion, supra ¶ 618, Google’s September 2020 changes to the SQR hobbled advertisers’ ability to identify appropriate negative keywords, exacerbating the issues surrounding the expanding match types. Tr. 5482:18–5483:9 (Jerath (Pls. Expert)) (“[T]hat’s the related problem that, as an advertiser, I’m being opted into these expansions, [] and I’m matching some of them. But in my search query report, I don’t have data on all the queries I’m being matched to.
So how am I supposed to come up with negative keywords if I don’t even know what all am I being matched to.”); Des. Tr. 157:14–17, 157:20–158:4, 158:20–160:22 (Alberts (Dentsu) Dep.) (match type expansions meant that its teams “needed to rely on those search query reports more frequently” to identify whether clients were subject to unwanted or inappropriate matches, so that it could “take action by either blocking keywords, pausing keywords, or incorporating negative keywords which will suppress the instances . . . match[es] to a particular term.”).
1184. Google’s changes to SQR reporting and match-type expansions reflect a slow erosion over time of advertisers’ ability to control and manage their ad spend, with the effect of higher prices to advertisers. Tr. 5494:11–5495:7 (Jerath (Pls. Expert)) (“Google restricts what advertisers can know about their own ad spend []. So you’re spending money but don’t know where it’s going. This was the example of [] removing data that was previously provided in the search query reports, and along with making things more difficult to opt out of through the negative keywords.
So all of this together, as I’d explained, leads to thicker auctions, and higher ad prices and unwanted ad spend by advertisers. So higher ad price is the important point here”; discussing UPXD103 at 45); UPX0511 at -619 (Amazon internal document outlining the “[s]low erosion of channel control” following Google’s changes over the years, including to keyword matching and SQRs); id. (“Google has removed and reduced the granularity of reporting data they make available”); Des. Tr. 160:18–162:20 (James (Amazon) Dep.) (discussing UPX0511).
d) Google Harmed Advertisers By Removing “Average Position” Metrics From The SQR
1185. Before September 2019, Google’s SQRs included “average position” metrics, which would tell the advertiser the average position on the SERP in which each of their ads was showing (i.e., between 1 and 4), as well as how a given ad’s position trended over time. Tr. 5175:14–5176:2, 5200:17–5201:2, 5202:3–5 (Booth (The Home Depot)) (describing former ad position metrics); Des. Tr. 152:14–153:7 (James (Amazon) Dep.) (describing ad position metrics previously available in SQRs); UPX8037 at .001 (A Google Ads Help webpage on ad position metrics explained that “[a]verage position reflect[ed] the order that your ad appear[ed] versus the other ads in the ad auction.”).
1186. Google has removed and replaced average position metrics with a less granular and less useful metric, again reducing the quality of Google’s reporting and impeding advertisers’ ability to assess their ad spend and manage their costs. UPX0511 at -619 (“In addition, Google has removed and reduced the granularity of reporting data they make available (position indicators, reduction in quality of SQR data, etc.).”); Des. Tr. 226:7–227:7 (Alberts (Dentsu) Dep.) (Average ad position is a common metric used by advertisers to optimize spend, and the solutions introduced after the deprecation of average position are less useful.).
1187. Advertisers used the average position metrics as an input when modeling the impacts of bidding changes and optimizing their bids. Tr. 5176:3–11 (Booth (The Home Depot)) (“It helps us understand where we show up on the search results page . . . .”); Des. Tr. 150:24– 151:19, 152:14–153:7 (James (Amazon) Dep.) (Average page rank could be an input into bidding optimization.). The advertiser could identify if there were opportunities to appear higher on the SERP and modify their bids accordingly. Tr. 5176:6–11 (Booth (The Home Depot)). Or, the advertiser could determine that they are already bidding at a sufficient level for where they want their ad to appear. Des. Tr. 40:10–21 (James (Amazon) Dep.).
1188. Alternatively, the advertiser could determine that it would be more cost-effective and yield a better return on investment for their ad to appear in a lower position on the SERP— e.g., in the second ad slot rather than the first one—and therefore lower their bid to aim for a lower average position. Tr. 5176:12–5177:1 (Booth (The Home Depot)) (“Not always would we want to be in the very absolute position of 1.0. With that typically comes a cost per click premium.
And so if we would identify that we’re always in a position of 1.0 or 1.1, it’s like, ooh, maybe we don’t need to be that aggressive in our bidding and we could lower our costs per clicks, lower our bid to essentially generate a stronger return on advertising.”); Des. Tr. 40:22–25, 152:14–153:7 (James (Amazon) Dep.) (“[With average position metrics,] we could understand as, for example, an ad moves from position one to position two in the search results, if we see a higher ROI in position two, which is entirely possible, versus being in position one.”).
1189. In November 2018, Google introduced “impression metrics,” which report only what percentage of an advertisers’ ads appear in one of the top positions on a page and what percentage appear in the first position. UPX8037 at .001 (Google Ads Help webpage on ad position metrics). Google initially continued to provide average position metrics alongside impression metrics, but in September 2019, Google ceased providing average position metrics. UPX8031 at -085; DX2021 at .001. Advertisers can no longer tell the allocation of their ads between positions 2, 3, and 4. Tr. 5175:14–5176:2, 5221:1–5222:2 (Booth (The Home Depot)).
1190. Google’s new impression metrics, however, did not provide the same information that the average position metrics did. Des. Tr. 152:8–153:7 (James (Amazon) Dep.) (“[W]e end up with a relative metric in terms of how often our ad was showing up in the top position, but doesn’t give us a strong signal in terms of where it might be otherwise on the page at the time.”); Tr. 5221:2–5222:1 (Booth (The Home Depot)) (The new impression metrics “[do] not give us the absolute number of the position or average position, no.”).
They were thus less helpful to advertisers, impacting their ability to optimize ad spend. Des. Tr. 153:8–13 (James (Amazon) Dep.) (Google’s removal of average position metrics from SQRs impacted Amazon’s ability to optimize ad spend); Tr. 5177:2–15 (Booth (The Home Depot)) (“[T]he more information that we get, I think the more intelligent we can be with our ad buying. I don’t know if it was catastrophic -- it certainly wasn’t catastrophic. But the types of insights that we would get, we wouldn’t have the same specificity.”).
1191. Amazon—which spends more than $[redacted] billion per year on Google Search Ads— pushed back on the removal of average position metrics from SQRs, but Google nevertheless proceeded with the change. Des. Tr. 41:1–22, 151:20–152:7 (James (Amazon) Dep.) (discussing UPX0061 and later referring to his team’s push back); UPX0061 at -437 (internal Amazon email with notes for an upcoming Google meeting: “[T]hey are continually reducing the quality of reporting data we are getting. I am continually arguing with them to give us visibility to the metrics that allow for us to understand our business drivers.”).
1192. In contrast to Google’s SQR changes, Microsoft provides richer reporting on Bing Search Ads. Des. Tr. 41:23–42:19 (James (Amazon) Dep.) (“We have more information, richer reporting information that we get from Bing relative to what we see from Google.”).
Continue Reading Here.
About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings.
This court case retrieved on April 30, 2024, storage.courtlistener is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction.