What the FTC got wrong in the Google antitrust investigation

When the FTC’s economists misread the search market

The FTC’s 2011–2012 antitrust investigation into Google ended with the agency closing the case. A decade later, Politico’s release of 312 pages of internal FTC memos offers a rare look at how that decision was made. For anyone working in tech, the documents are troubling for a specific reason: the arguments that won—those urging the FTC to stand down—are built on a shaky understanding of how the search industry actually works.

The investigation split the FTC internally. The Bureau of Competition (BC) argued that antitrust action was warranted. The Bureau of Economics (BE) argued that the case should be dropped. The BC memo is defensible, even if reasonable people could disagree with its conclusions. The BE memo is another matter. It contains fundamental errors in core arguments, and there is no public sign that anyone at the FTC noticed or pushed back. The memos from directors and higher-ups suggest leadership gave the BE memo at least as much weight as the BC memo, which implies a gap in the FTC leadership’s understanding of the tech industry.

The case for action

The BC memo’s executive summary lays out a straightforward antitrust theory. Google held monopoly power in three U.S. markets: horizontal search, search advertising, and syndicated search and search advertising. The memo identified four areas of potentially anticompetitive conduct, though it did not recommend pursuing all of them.

• Search bias: The BC staff did not recommend action over Google’s preferencing its own content in results while demoting rivals. It was a close call, with unfavorable case law on anticompetitive product design and strong efficiency justifications.
• Content scraping: BC recommended condemning Google’s practice of scraping content from vertical rivals for its own vertical products as a conditional refusal to deal under Section 2. The theory was that prior voluntary dealing was mutually beneficial, that Google threatened to remove rivals from general search to force them into allowing content use, and that the likely effect was to reduce vertical sites’ incentives to invest in R&D.
• Ad campaign restrictions: BC argued that contractual limits on automated cross-management of ad campaigns should be condemned under Section 2. These restrictions prevented advertisers from using their own data, raised transaction costs, and degraded the quality of Google’s search rivals. Google’s efficiency justifications appeared pretextual.
• Exclusionary syndication agreements: BC recommended condemning exclusive agreements for syndicated search and search ads. The effects on publishers were modest, but the agreements denied scale to competitors—particularly Bing—and created significant long-term barriers to entry.

The memo also outlined potential remedies: an opt-out for snippets from Google’s vertical properties, removal of problematic contractual restrictions from license agreements, and an end to exclusive search syndication deals.

BC staff acknowledged risks to the case. Google could argue that Microsoft’s most efficient distribution channel is bing.com and that any scale Microsoft gained would not materially improve Bing’s competitive position.

The mobile supplement

A supplemental BC memo on mobile argued that Google dominated mobile search through exclusivity agreements, in a market that was growing rapidly. Google’s internal documents showed mobile search growing from 9.5% to 17.3% of searches in 2011, and both Google and Microsoft expected mobile to surpass desktop in the near future. BC staff cited Google leadership acknowledging that Google could unilaterally reduce revenue share as evidence of monopoly power.

The BC staff did not paint Google as a simple villain. They acknowledged that many of Google’s actions benefited consumers, but argued the company was simultaneously engaging in tactics that harmed vertical competitors and helped entrench its monopoly over search and search advertising.

The BE staff disagreed, and disagreed broadly. In almost every instance, the BE memo argued that the relevant market was not properly defined, was not important, or was competitive—and that Google’s conduct was therefore procompetitive.

When the FTC’s own evidence contradicted its antitrust case

Politico’s release of FTC documents from the Google antitrust investigation offers a rare look at how the Bureau of Economics (BE) and Bureau of Competition (BC) staffs framed their cases. Even a casual reading shows the BE memo repeatedly stretched data and relied on implausible assumptions to argue against action. The deeper problem: FTC directors appear to have weighed those weak arguments as seriously as the BC staff’s evidence-backed case.

Mobile dismissed as a sideshow

The BE memo’s central claim was that mobile was a small, unimportant market — only 8% of overall queries and an even smaller slice of search ad revenue. To support the idea of robust mobile competition, the memo pointed to BlackBerry and Windows Mobile. But between the start of the FTC investigation and the memo’s writing, BlackBerry’s share had fallen from roughly 14% to 6%, part of a long-term decline with no sign of reversal. Windows Mobile slid from about 6% to 4%. In a market with strong network effects, treating platforms with low and falling share as competitive counterweights was already questionable.

At the time, the industry consensus was the opposite. Google had already pivoted to a “mobile first” strategy, and other large tech companies had done the same. The redesigns that degraded desktop experiences in service of mobile-first development were common sources of user complaints — a sign that companies understood where the market was heading.

Selective use of comScore data

The BE memo argued that Yahoo and Bing provided robust, stable competition — citing a combined “steady” 30% U.S. market share and claiming Microsoft’s query volume was growing faster than Google’s. It even went so far as to suggest Microsoft and Yahoo together had more active search users than Google, measured by MAU.

These claims relied on comScore data, but the memo’s use of that source was inconsistent. When comScore numbers made Microsoft look weak — as in syndicated search — the authors said the data was inaccurate and shouldn’t be used. When the same data painted an implausibly favorable picture for Microsoft, they leaned on it without explanation.

The underlying argument was that because some users occasionally used Yahoo or Bing — perhaps once a month versus a thousand Google searches — there could be no meaningful barrier to switching. Anyone who has worked on growth understands that converting a lightly engaged user into a heavy one is generally harder than converting a brand-new user. Light engagement doesn’t demonstrate that competition is thriving.

What the memo omitted: the fixed costs of running a search engine are enormous, Yahoo had effectively outsourced search to Bing, and Microsoft was subsidizing Bing at roughly $2B per year in a bet most industry observers doubted. Short of that subsidy, the likely outcome was a significant share drop — which is exactly what happened.

Defaults treated as worthless

Another remarkable position in the BE memo was that search defaults were essentially meaningless because users could switch with “a few taps” in “a few seconds.” The memo described these as “trivial switching costs.”

The most direct evidence against this claim is what Google actually paid for default status. A 2023 antitrust case revealed Google paid Apple $26.3B in 2021 for the privilege of being the default search engine. A rough calculation: if that payment were pure profit to Apple, it could account for roughly $776B of Apple’s market cap. From Google’s side, with a similar P/E ratio, the company was effectively giving up something on the order of $722B of its own valuation. There is no reasonable adjustment that makes those numbers trivial.

Numbers closer to the FTC investigation period tell the same story. TechCrunch reported in 2013 that Google paid Apple $1B per year for search status, and a lawsuit later confirmed the $1B payment for default status in 2014. That’s still a substantial sum — and it directly contradicts the memo’s dual claims that mobile was unimportant and defaults didn’t matter.

It’s even stranger given the BE memo’s insistence elsewhere that the BC staff should rely on data rather than abstract reasoning. When the evidence pointed toward meaningful default effects, the memo’s authors ignored the obvious source — the payments Google was making — and substituted casual reasoning.

A static model of dynamic markets

The BE memo also argued that Google could not seriously harm vertical competitors like Yelp or TripAdvisor because Google drove only 10% to 20% of their traffic. The implicit model was static: if Google removed or downranked Yelp listings, Yelp would lose at most that fraction of traffic, permanently.

In a dynamic market, the expected outcome is different. If Google persistently siphons user traffic to its own local results — while simultaneously presenting Yelp’s current users with an alternative bundled into products they already use — the long-run result is that Yelp loses its growth channel and eventually withers. That is what happened. Yelp now trades at a value around $2B with an unimpressive P/E ratio, because investors don’t expect it to recover against Google’s search and maps dominance.

This wasn’t a secret at the time. A former Google colleague working on local features told me the expected outcome was to cripple Yelp’s business, and that Yelp could not counter it because of Google’s market power in search and maps. The BE memo’s confidence that cutting off a company’s “air supply” would have no significant effect mirrors the logic that Microsoft allegedly applied to Netscape — except the FTC staff treated it as a reasonable position.

Marketing statements treated as evidence

The BE memo repeatedly leaned on claims that were obviously not statements of fact. One example: the memo quoted Microsoft CEO Steve Ballmer’s press release announcing the Yahoo search agreement — “This agreement with Yahoo! will provide the scale we need to deliver even more rapid advances in relevancy and usefulness” — as if CEO marketing language were meaningful evidence. This kind of pablum accompanies nearly every partnership and acquisition. Regulators should recognize it as marketing, not data.

The asymmetry in evidentiary standards

Across the memos, a pattern emerges. The BC staff referenced interviews and internal documents from major tech companies, including the hyperscalers. The BE staff, with access to the same materials, argued that mobile was unimportant, that defaults were worthless, and that cutting off rivals’ primary user-acquisition channel was inconsequential.

Directors who weighed the two cases sided with the BE memo — one strongly, one moderately. Their comments thanked both staffs for “outstanding work” and noted the case was “close” on four areas, without any indication that the BE memo’s more implausible claims were scrutinized or tested against available evidence.

At times, the BE memo argued that the BC staff’s reasoning was anecdotal or speculative when it suggested barriers to competition. But when the data showed Google’s conduct created barriers — as with default payments — the BE authors discarded that data and relied on weak abstract reasoning instead. If the evidentiary bar for pursuing an antitrust case was meant to be higher, the memo could have said where the evidence fell short without making its own unsupported assertions.

An observer with industry knowledge who read only the BE memo would conclude mobile was irrelevant, defaults were trivial, and vertical competitors had nothing to fear. A reader of the BC memo alone would conclude the opposite. The FTC had access to internal documents from the companies involved and could have checked how much Google was paying for default status — one director’s memo even suggested someone ought to verify that number. The investigation was dropped shortly afterward, with no evidence that anyone did.

Oddities from the investigative record

Beyond the core analytical disagreements, the memos contain a number of notable details about both the investigation itself and Google's conduct.

Between the approval of compulsory process in June 2011 and the publication of the BC memo in August 2012, staff received 9.5 million pages of documents across 2 million individual documents. The staff reported reviewing "many thousands of these documents," meaning only a small fraction of the total material was ever examined.

Prior to the FTC investigation, several lawsuits raising similar claims had already been dismissed. In SearchKing v. Google, the court ruled that Google's rankings were constitutionally protected opinion, meaning even malicious manipulation of results would not create liability. In Kinderstart v. Google, part of the ruling held that Google search was not an essential facility for vertical providers such as Yelp, eBay, and Expedia. The memos also contain extensive discussion of Verizon v. Trinko and Aspen Skiing Co. v. Aspen Highlands Skiing Corp.

At the time the BC memo was written, 96% of Google's $38 billion in revenue came from advertising, mostly search ads. The memo argued that other advertising formats besides social media had limited growth potential — a claim that looks clearly wrong in retrospect. YouTube alone generated $28.8 billion in ad revenue in 2021, and significant video ad spending bypasses platforms entirely. The #137th largest streamer on Twitch was reportedly offered $10 million per year to stream online gambling for 30 minutes a day. Even at the time, strong signals indicated video would become a major advertising market, though those same signals also suggested Google would dominate it.

The BC memo generally overstated both the expected primacy of search ads and how distinct a market search advertising was, claiming other online ad spend was not a substitute in any way and was instead a complement. One could reasonably argue search ads form a somewhat distinct market with low elasticity of substitution, but the memo pushed this much further. This may have been a reaction to Person v. Google, where Judge Fogel criticized the plaintiff's market definition for failing to distinguish a "search advertising market" from the broader internet advertising market. Still, as a factual matter, the argument appears dubious.

Vertical integration and the onebox fight

The BC memo's treatment of Google's integrated products like local search and product search (formerly Froogle) is particularly detailed. The memo claimed that if Google treated its own properties the way it treated other websites, those products would not be ranked at all — and that Google artificially placed its own vertical properties above organic results.

The webspam team initially declined to include Froogle results, saying the results were exactly the kind of spammy content Google normally removes from the index: "[o]ur algorithms specifically look for pages like these to either demote or remove from the index." Bill Brougher, product manager for web search, said, "Generally we like to have the destination pages in the index, not the aggregated pages. So if our local pages are lists of links to other pages, it's more important that we have the other pages in the index." The webspam team was overruled and the results were inserted. The ads team then complained that the lower-quality results would cause a loss of $154 million per year. The response tracked the BC memo's argument about scale and the cost of depriving competitors of it:

We face strong competition and must move quickly. Turning down onebox would hamper progress as follows - Ranking: Losing click data harms ranking; [t]riggering Losing CTR and google.com query distribution data triggering accuracy; [c]omprehensiveness: Losing traffic harms merchant growth and therefore comprehensiveness; [m]erchant cooperation: Losing traffic reduces effort merchants put into offer data, tax, & shipping; PR: Turning off onebox reduces Google's credibility in commerce; [u]ser awareness: Losing shopping-related UI on google.com reduces awareness of Google's shopping features

Normally, CTR is a strong ranking signal, but applying it here would have ranked Google's own vertical properties poorly. Instead, Google used the presence of competing vertical websites to automatically boost its own properties above those competitors — when a comparison shopping site was relevant, Google Product Search would appear above it, and when a local search site like Yelp or CitySearch was relevant, Google Local would appear at the top of the SERP.

Google also integrated Yelp content into Google Places. When Yelp objected, Google threatened to ban Yelp from traditional search results, and further threatened to ban any vertical provider that refused to allow its content to be used in Google Places. Marissa Mayer testified that it was technically extraordinarily difficult to remove Yelp from Google Places without also removing Yelp from organic results. When Yelp sent a cease and desist letter, Google removed Yelp results immediately — suggesting the difficulty was overstated. Google then claimed it was infeasible to remove Yelp from Google Places without removing Yelp from the "local merge" interface on the SERP. BC staff believed this too was false, and Mayer later admitted in a hearing that it was false, saying Google was concerned about the consequences of allowing sites to opt out of Google Places while remaining in "local merge." A similar sequence played out with Amazon and product search. The BE memo's counterargument was that Google traffic was "very small and not statistically significant."

The BC memo argued these actions reduced the incentives of companies like Yelp, CitySearch, and Amazon to invest, and discouraged new entrants. That appears true. Around the time of the FTC investigation, founders and VCs had already moved away from funding companies like Yelp because it was understood Google could seriously cripple any similar company in that space by cutting off its traffic.

AdWords, syndication, and missing notes

The appendix discussion of AdWords API restrictions covers provisions that disallow programmatic porting of campaigns to other platforms such as Bing. Notably, Google appears to have been aware of the legal sensitivity: meeting notes and internal documentation on the topic are unusually incomplete. The most informative written record BC staff could find was a message from Director of PM Richard Holden to SVP of ads Susan Wojicki: "We didn't take notes for obvious reasons hence why I'm not elaborating too much here in email but happy to brief you more verbally."

The memos also discuss Google's exclusive and restrictive syndication agreements. Google claimed it was unaware of the terms and conditions in its standard online service agreements, which contained a "preferred placement" clause that many parties viewed as a de facto exclusivity agreement. When FTC staff questioned Google's VP of search services about this, he said he was not aware of the term. Google subsequently sent a letter to Barbara Blank of the FTC stating it was removing the preferred placement clause from the standard online agreement.

The BE memo cites most customers' lack of concern about these restrictive terms as evidence they were not a problem. In retrospect, that unconcern appears to reflect a failure to understand the stakes of online commerce. Only a small number of Google's largest and most sophisticated customers, such as Amazon and IAC, found the terms concerning — and their concern was that Google's restrictions would increase its dominance over Bing/Microsoft, allowing Google to dictate worse terms later.

Best Buy viewed its website and the web as a way for customers to find pre-sale information before entering a store. Walmart viewed the web as an extension of brick-and-mortar retail. Both retailers, previously in much stronger positions than Amazon, fell substantially behind in online and overall profit. Walmart eventually realized its error, acquiring Jet.com for $3.3 billion in 2016 and funding serious technology work internally. Since then, Walmart has posted roughly 30% CAGR in online net sales since 2018 — but after taking two decades to mount a serious response to Amazon, it remains solidly behind in online retail. Best Buy has still not mounted an effective response after three decades.

When the BE memo refers to these customers as sophisticated, that is relative to retailers whose leadership did not understand the internet. In the tech industry at the time, one did not need to find a particularly sophisticated individual to find someone who understood what was happening. It was generally understood that retail revenue and even more so retail profit would move online.

Scale, click data, and the quality of search

Both memos contain lengthy discussions of search and scale. The BC memo's rough argument is that multiple markets — search and ads — have significant scale effects on product quality. Google's own documents acknowledge this "virtuous cycle": more users enable better ads, which generates more revenue; more scale provides more data to improve search results, which drives user growth. For search specifically, the BC memo claims click data is highly important and that more data enables better results.

The BE memo raises two objections. First, that the importance of click data is "contrary to the history of the general search market." Second, that it is "also contrary to the evidence that factors such as the quality of the web crawler and web index; quality of the search algorithm; and the type of content included in the search results [are as important or more important]."

The first argument is essentially: Google used to be smaller, had sufficient click data at that size, and therefore current-scale competitors with similar data should be fine. This ignores that Google's earlier success came when the market was less mature and no one was producing a better product. It is especially weak in markets with a virtuous cycle between market share and product quality, like search.

The second argument is also strange, even setting aside technical knowledge. It resembles claiming that because a car needs a left front tire and a right rear tire, the right front tire isn't important. Understanding search makes the argument even less plausible. Scale and click data directly feed algorithm development; the search algorithm is not a separate input. As someone who has worked on search indexing, it is also difficult to agree that indexing matters as much as ranking. Indexing and crawling are easier and less important problems than ranking. This was generally understood at the time, and the BE memo's authors, having conducted numerous staff interviews, should have known it.

The example of Cuil is instructive. Cuil built a larger index than Google, which was not a trivial achievement, but the company failed because a large index without strong ranking is worth little. While it is technically true that good ranking with a poor index is also worth little, this situation rarely occurs in practice — a company competent enough to build a strong ranker will as a matter of course have adequate indexing and crawling.

The BC memo's case is stronger: increased scale greatly improves search quality; the extra data Bing gained from Yahoo materially improved quality and CTR; further scale should be expected to produce continued returns; the costs of creating a competitor are high — Bing was said to be losing $2 billion per year at the time and spending $4.5 billion per year developing algorithms and building physical capacity; and Google's potentially anticompetitive actions disadvantaged Bing relative to a counterfactual. The same general case is made for ads.

Yet the BE memo's case is correct in spirit: Microsoft could have taken actions it did not take to compete far more effectively in search, and one could argue the FTC should not be in the business of rescuing a company from competing ineffectively. The positions the BE memo takes are extremely weak, but the decision at the FTC appears to have turned more on philosophy than on the technical arguments in either memo.

The Road Not Taken in 2012

Had the FTC pressed forward with its antitrust investigation into Google in 2012 rather than closing it, the practical outcomes might have been modest at best. The most likely scenario, given court attitudes at the time, would have been a nominal settlement — a fine representing a tiny fraction of Google's profits from the challenged conduct, or a consent decree requiring the company to stop specific practices while retaining its market share. History offers plenty of parallels. The EU's Android tying case nominally succeeded but produced a remedy that was easily circumvented; as Cristina Caffara of the Centre for Economic Policy Research put it, regulators told Google "don't do it again, bad dog," and the company responded with a pricing structure that effectively preserved its default search deals. Similar EU actions on shopping verticals arrived years too late to matter.

The contrast with Russia is instructive. When Yandex brought its case over mobile search defaults in 2015, it still held marketshare in the high 30% range. The resulting choice screen quickly helped Yandex surpass Google. By the time the EU imposed a similar screen in 2018, Bing held roughly 3.6% of the European market — a choice screen in a market that lopsided offers little practical change. The window for cheap intervention in winner-take-most markets is narrow, and the BE memo's caution about downstream harms may have helped close it entirely.

The Microsoft Precedent

The history of the Microsoft antitrust saga suggests that even successful action can arrive too late for the intended beneficiaries. The 1998 DoJ suit over browser bundling targeted Netscape's destruction; by the time the case settled in 2002, Netscape was effectively dead. Conservative economists of the era, including Milton Friedman, warned that antitrust action itself would trigger a regulatory spiral. Friedman wrote that Silicon Valley companies calling for action against Microsoft were committing "suicidal impulse," predicting "a continuous increase in government regulation" and the "bureaucratization" of the industry.

That prediction proved wrong on both counts. The two decades following the Microsoft case saw relatively light antitrust enforcement, and the case may have indirectly enabled Google's survival. Internal Microsoft discussions in Google's early days included proposals to redirect users away from google.com or display malware-style warnings about the site. Gene Burrus, Microsoft's lawyer at the time, said the company declined to pursue these ideas out of fear of further antitrust scrutiny. People at both companies who were interviewed about this period believe Microsoft would have killed Google had it attempted such measures.

What Biden-Era Appointees Might Have Done

If the FTC and DoJ in 2012 had been led by officials sharing the philosophy of Biden's appointees — Lina Khan and Jonathan Kanter — the investigation might have proceeded. Kanter and Khan's approach is less concerned with whether courts will uphold novel theories and more focused on enforcing the laws on the books. The Obama-era appointees who let the investigation lapse largely rejected this view; GMU economist Josh Wright, an Obama FTC commissioner, authored a rebuttal to Khan's influential "Amazon's Antitrust Paradox" titled "Requiem for a Paradox."

Biden's appointees have pursued cases against Meta and Microsoft that Obama-era enforcers likely would not have brought, and many have failed in court. Had similar action been taken against Google in 2012, the likely result would have been a settlement with little practical effect. Microsoft would almost certainly have scaled back Bing investment before any decision landed, given the multi-year timeline of such cases.

A Better Outcome Was Possible

A more aggressive antitrust posture in 2012 might have changed the competitive landscape in ways that matter well beyond Bing versus Google. The BE memo leans heavily on the idea that intervention risks cascading harms, but that argument cuts both ways. Microsoft's dominance in adjacent markets has shaped outcomes elsewhere: Teams defeated Slack not because it was better but because Microsoft could bundle and leverage its position. Comments from industry observers suggest Slack was effectively forced into acquisition because competing with Teams given Microsoft's enterprise dominance was untenable. Had antitrust constraints existed on such leverage, independent products and new startups in enterprise tools might have flourished.

Effective remedies are extraordinarily difficult to craft, however. A 2012 choice screen for browsers and search would have presented real problems for Mozilla, which relies on Google payments for default status to fund its operations. Opera's history shows that even a superior paid browser cannot compete with free alternatives. Regulatory actions also carry unintended consequences beyond their immediate targets. The Bundeskartellamt's restrictions against Facebook over data use in Germany illustrate the jurisdictional limits of such remedies: Facebook limits data processing only when it detects users are German, leaving global ML training largely unaffected. For American regulators, similar concerns about imposing competitive drag on domestic firms relative to foreign rivals would weigh on any ambitious remedy.

The speculative best case for antitrust advocates — timely action that preserved meaningful competition in search, verticals, and enterprise software — is difficult to envision as a realistic outcome. The legal and political constraints were substantial, and the BE memo's arguments, however flawed in places, identified real risks. At minimum, the record suggests that delayed or diluted enforcement has costs of its own, and the caution that prevailed in 2012 was not a neutral choice.

Bringing technical expertise into the room

A recurring theme in the FTC's investigation is that the people crafting the arguments simply didn't have enough technical fluency to tell plausible claims from implausible ones. Looking at the publicly available director memos from the 2011–2021 investigation, arguments from the BE memo that wouldn't pass the sniff test for someone with a tech background appear to have been taken seriously. Having technically literate people involved not just in investigations but also in crafting remedies and regulation would likely have changed the outcome.

Consider one EU remedy that Cristina Caffara noted was immediately worked around by Google. To people in tech, that workaround looks like a clever "hack," not an act of defiance. This kind of system-beating behavior has a long history in the field, going back to when it was just physics and electrical engineering.

There's a well-known example from Paul Graham's 2010 essay on founders. Under the section titled "Naughtiness," he wrote:

Though the most successful founders are usually good people, they tend to have a piratical gleam in their eye. They're not Goody Two-Shoes type good. Morally, they care about getting the big questions right, but not about observing proprieties. That's why I'd use the word naughty rather than evil. They delight in breaking rules, but not rules that matter. This quality may be redundant though; it may be implied by imagination.

Sam Altman of Loopt is one of the most successful alumni, so we asked him what question we could put on the Y Combinator application that would help us discover more people like him. He said to ask about a time when they'd hacked something to their advantage—hacked in the sense of beating the system, not breaking into computers. It has become one of the questions we pay most attention to when judging applications.

That same spirit shows up inside companies all the time. At Google, engineers built a system to reduce travel costs by computing a baseline expected price for flights and giving employees a credit for flights under that baseline, which could be used to upgrade future accommodations. Compared to rigid expense limits at more traditional companies, this was genuinely nicer for employees. Then some employees started optimizing aggressively—routing trips through locations that were highly optimizable to rack up credits for first-class upgrades and nicer hotels. When talking to people in traditional industries about this, they're often horrified that these employees weren't censured or fired. But at Google, people generally found it admirable; it exemplified the hacker spirit.

Courts, regulators, and legislators have not been prepared for the vigor, speed, and delight with which tech companies hack the system. This goes back at least two decades of antitrust in tech. There's precedent for bringing tech people to the other side of the table—it happened in the big Microsoft antitrust case—but there are serious structural problems in doing so.

The incentive problem

Tech folks who are very good at hacking systems and who want to be employed at big companies frequently make seven figures (or more) annually. That sum isn't likely to be rivaled by an individual consulting contract with the DoJ or FTC. In the Microsoft case, the technical group was managed by Ron Schnell, who was taking a break after his third exit, but people like that are relatively few and far between. There are people who refuse to work at big companies for moral reasons or because they dislike corporate politics, but most of them haven't spent enough time at big companies to understand how they actually operate—making them the wrong people for the job even if they're great engineers and hackers.

At an antitrust conference, a speaker noted that collaboration between the legal and economics communities was a great boon for antitrust work. Notably absent from both the speech and the conference were practitioners from industry. The conference had the feel of an academic gathering; you might see CS academics at one eventually, but many policy-level discussions are outside their area of interest. For instance, the BE memo used MAU numbers to argue that switching costs were low—a claim that almost no CS academic would be equipped to evaluate.

That lack of technical collaboration has consequences beyond policy discussions. When people speculated about motives, they often made unwarranted assumptions. A speaker described an executive's reaction to some hack—high-fives and celebration—and inferred contempt for lawmakers and the law. That's not supported by the evidence. That celebration looks exactly like what you'd see after someone at Google figured out how to get upgraded to first class "for free" on almost all their flights. It doesn't indicate contempt or disdain at all.

The incentive problem extends well beyond getting tech people into antitrust discussions. Capitol Hill staffers from the time generally believe the primary factor that scuttled the FTC investigation was Google's lobbying. Google and other large tech companies outspend entities interested in increased antitrust scrutiny by a wide margin.

There's also a revolving-door problem in civil service. The lead of the BC investigation and first author on the BC memo are now Director and Associate General Counsel of Competition and Regulatory Affairs at Facebook. Whatever their personal motivations, the offer must be compelling. Even putting aside the pay, someone who strongly believes in antitrust enforcement faces a grim choice: stay at the FTC and lead another investigation where a well-argued memo gets ignored when a big tech company pours lobbying money into D.C., or watch an EU-style "choice screen" remedy arrive too little and far too late—or see a consent decree from an Android Play Store untying case become useless in about five minutes when an engineer figures out a hack, seven years after the investigation started. At Facebook, at least, you can nudge the company toward what you think is right and have some impact on how it treats consumers and competitors.

Among people in tech, it's common to hear "I'd never work at company X for moral reasons." That's a fine position, but almost everyone who takes it ends up at a much smaller company with almost no impact on the world. If you want to make a difference from a moral standpoint, you're more likely to succeed by working from the inside or by helping a smaller direct competitor become more successful.

Appendix: non-statements

This is analogous to the "non-goals" section of a technical design doc, but weaker—a non-goal in a design doc is often a positive statement that implies something that couldn't be inferred otherwise, whereas these statements don't add information.

• Antitrust action against Google should have been pursued in 2012
  • Not that anyone should care what my opinion is, but if asked at the time, I would've said "probably not." The case for antitrust action seems stronger now and the case against weaker, but you could still mount a fairly strong argument against antitrust action today.
  • Even if you believe that, all else equal, antitrust action would've been good for consumers, it's still not obvious that Google and other tech companies are the right target as opposed to—for example—Visa and Mastercard's dominance of payments, hospital mergers leading to increased concentration that's hurt consumers and workers, Ticketmaster's dominance, or areas where regulation specifically protects firms, such as shipping (exempt from the Sherman Act) or car dealerships (which have special legal protections in many states).
• Weaker or stronger antitrust measures should be taken today
  • I haven't spent enough time reading up on the legal, political, historical, and philosophical background to have an opinion on what should be done, but I know enough about tech to point out a few errors I've seen and to call out common themes in them.

BC Staff Memo

By "Barbara R. Blank, Gustav P. Chiarello, Melissa Westman-Cherry, Matthew Accornero, Jennifer Nagle, Anticompetitive Practices Division; James Rhilinger, Healthcare Division; James Frost, Office of Policy and Coordination; Priya B. Viswanath, Office of the Director; Stuart Hirschfeld, Danica Noble, Northwest Region; Thomas Dahdouh, Western Region-San Francisco, Attorneys; Daniel Gross, Robert Hilliard, Catherine McNally, Cristobal Ramon, Sarah Sajewski, Brian Stone, Honors Paralegals; Stephanie Langley, Investigator"

Dated August 8, 2012

Sizing Up the FTC’s Google Case

After a 19-month investigation that produced more than nine million pages of documents and testimony from Google executives, the FTC staff reached a nuanced conclusion. The inquiry identified four areas of potential anticompetitive conduct, plus a separate track for mobile. Staff recommended action on three of those four areas, but explicitly declined to recommend action on search bias — the most publicly controversial question.

Where Staff Saw Violations

The clearest case, in staff’s view, involved Google’s scraping of content from vertical rivals. The theory: Google first engaged in mutually beneficial voluntary dealings with vertical sites, then threatened to remove their content from general search results unless those sites allowed Google to use their content in Google’s own vertical products. Staff said this amounted to a conditional refusal to deal under Section 2, with the natural and probable effect of diminishing rivals’ incentives to invest in research and development.

Staff also recommended condemning Google’s contractual restrictions on automated cross-management of ad campaigns. Those restrictions, staff argued, limited advertisers’ ability to use their own data, raised transaction costs, reduced innovation, and degraded the quality of Google’s rivals in search and search advertising. Google’s efficiency justifications for the restrictions appeared, on inspection, to be pretextual.

On exclusionary syndication agreements — deals that made Google the default search provider on partner sites — staff saw only modest anticompetitive effects on publishers themselves, but found the agreements competitively significant because they denied scale to Bing and raised long-term barriers to entry. Here, too, staff found Google’s efficiency rationales unpersuasive.

What Staff Let Go

The one area staff declined to pursue was the allegation that Google illegally preferenced its own content in search results while demoting rivals. The memo describes the question as a close call, with case law that is not favorable to claims of anticompetitive product design. Google’s efficiency justifications for displaying its own vertical content were strong, and the practice produced at least some benefit to users.

That decision stands in contrast to the theory of harm in the private litigation that surfaced during the investigation. In cases like Kinderstart.com LLC v. Google, Inc. and SearchKing, Inc. v. Google Tech., Inc., plaintiffs alleged that Google unfairly demoted or manipulated their rankings. Those cases were dismissed — in the SearchKing case, the court held that Google’s rankings are constitutionally protected opinion, and in Kinderstart, the court rejected the argument that Google’s search results constitute an essential facility for vertical websites.

AdWords-related lawsuits met a similar fate. TradeComet.com, LLC v. Google, Inc. was dismissed for improper venue, and Google, Inc. v. myTriggers.com, Inc. failed because the plaintiff could not describe harm to competition as a whole. In Person v. Google, Inc., the court criticized the plaintiff’s market definition, finding no basis for distinguishing a separate "search advertising market" from the broader market for internet advertising.

Remedies on the Table

Where staff did recommend action, the proposed fixes were narrowly tailored:

• Scraping: Google could be required to offer an opt-out for snippets such as reviews and ratings when those snippets appear in Google’s vertical properties, while retaining them in ordinary web search results and Universal Search. Alternatively, Google could be limited in how it uses content indexed from web search results.
• Campaign management restrictions: Google could be required to strip the problematic contractual restrictions from its license agreements.
• Syndication agreements: Google could be enjoined from entering into exclusive search agreements with syndication partners, and required to loosen restrictions on partners’ use of rival search ads.

The Context Around the Investigation

The FTC’s work ran parallel to, and coordinated with, other proceedings. The European Commission had been investigating since November 2010. In May 2012, Commissioner Joaquin Almunia signaled the EC’s possible intent to issue a Statement of Objections, citing concerns about Google’s favorable treatment of its own vertical services, its copying of third-party content, exclusivity agreements with publishers, and restrictions on cross-platform ad campaign management. Google denied infringement but proposed commitments to address those concerns.

At the state level, Texas led a multi-state working group that began its own investigation in June 2010, in close coordination with FTC staff.

The FTC’s own investigation relied on a compulsory process approved in June 2011, more than two million documents (9.5 million pages), records from the earlier Google-Yahoo and ITA reviews, and dozens of interviews with vertical competitors, advertisers, ad agencies, syndication partners, mobile device manufacturers, and carriers. Staff also conducted 17 investigational hearings with Google executives and employees.

Staff acknowledged several risks to its case. Google could argue that Microsoft’s most efficient distribution channel is Bing.com, and that any scale gained through syndication would be immaterial to Bing’s competitive position. Despite those risks, staff concluded that Google’s conduct had resulted, and would continue to result, in real harm to consumers and to innovation in online search and advertising.

The FTC’s Google Probe: A Technical Dissection of the Staff Report

The Federal Trade Commission’s investigation into Google’s search practices generated a voluminous staff report that examined five principal areas of alleged anticompetitive conduct. While the Commission ultimately declined to issue a complaint, the staff’s findings—and the internal Google documents they unearthed—paint a detailed picture of how the company leveraged its control over search to advantage its own properties. What follows is an engineering-focused review of the staff’s core arguments and the evidence supporting them.

Scale as a Defensible Moat

The report devotes considerable attention to the role of scale in search quality, arguing that query volume and advertising volume create powerful feedback loops that are difficult for competitors to disrupt.

Microsoft’s chief economist, Susan Athey, testified that insufficient search volume directly hampers Bing’s ability to run meaningful experiments. More traffic enables more simultaneous tests and faster completion times. Athey argued that relative scale—Bing being roughly one-fifth the size of Google—is what matters, not absolute size. Microsoft claimed that even a 5–10% increase in query volume would be “very meaningful,” pointing to the 2010 Yahoo data-sharing agreement, which it said improved its “auto suggest” click-through rate from 44% to 61% between July 2010 and September 2011.

Udi Manber, Google’s former chief of search quality, provided candid testimony on this dynamic: “If Microsoft had the same traffic we have their quality will improve *significantly*, and if we had the same traffic they have, ours will drop significantly. That’s a fact.” In a separate 2009 statement, Manber explained how click data directly shapes rankings: “If we discover that, for a particular query, hypothetically, 80 percent of people click on Result No. 2 and only 10 percent click on Result No. 1, after a while we figure out, well, probably Result 2 is the one people want.”

Google’s own documents are replete with references to a “virtuous cycle” among users, advertisers, and publishers. The company’s counterargument was that scale ceases to matter once a search engine reaches a certain absolute size, and that Bing’s additional queries would not “significantly improve” its quality. The staff report suggests this is a hard claim to reconcile with internal Google communications acknowledging the importance of traffic volume for experiment velocity and click-data feedback.

Preferencing Vertical Properties on the SERP

The most extensively documented allegation concerns Google’s deliberate promotion of its own vertical properties—shopping, local, flight, and others—within the main search results page (SERP).

Google began launching vertical properties around 2001, including Google News, Froogle (shopping), Image Search, and Groups. By 2005, internal strategy documents show the company viewing specialized vertical search engines as a potential “threat” to its dominance. A 2010 internal email stated: “Vertical search is of tremendous strategic importance to Google. Otherwise the risk is that Google is the go-to place for finding information only in the cases where there is sufficiently low monetization potential that no niche vertical search competitor has filled the space with a better alternative.”

The rollout of “Universal Search” in 2007 was the key technical turning point. Marissa Mayer, then leading the effort, described it in a 2003 memo as a redesign “SO that Google deliver[s] the most relevant information to the user on Google.com no matter what corpus that information comes from. This design is motivated by the fact that very few users are motivated to click on our tabs.”

While Universal Search was marketed as a user-experience improvement, internal documents suggest product triggers were often tuned to serve Google’s properties rather than user intent. One Google “Launch Report” described an algorithm that would trigger the Product OneBox on queries frequently searched on Google Shopping, “automatically plac[ing] the universal in position 4, regardless of the quality of the universal results or user ‘bias’ for top placement of the box.”

Jon Hanke, then head of Google Local, made the strategic intent explicit in an email to Mayer: “[T]he mandate has to come down that we want to win [in local] and we are willing to take some hits [i.e., trigger incorrectly sometimes]... results that are not web search results and that displace web pages are ‘OK’ on google.com.” He added that Google’s key strength was “Google.com real estate for the ~70MM of product queries/day in US/UK/DE alone.”

Eric Schmidt and Larry Page reportedly issued a mandate to push product-related queries as aggressively as possible. Google’s financial models in spring 2008 estimated that top placement of the Product Universal would cost the company $154 million per year in lost ad revenue on product queries. The Ads team requested reduced triggering frequency; the Product Universal team objected, arguing that lost traffic would harm ranking click-data, triggering accuracy, merchant cooperation, and user awareness of Google’s shopping features.

A particularly telling internal eBay click-through analysis (from January–April 2012) compared the performance of Google Product Search, natural web results, and eBay links at each SERP position. Across all positions, Google Product Search consistently had lower click-through rates than both the natural results and eBay’s links. At position 1, natural results saw 38% CTR, Google Shopping 21%, and eBay 31%. At position 5, the figures were 10%, 8%, and 10%, respectively.

Despite tracking CTR extensively for web rankings, Google did not use CTR to rank Universal Search results against web results. Mayer said this was “because it would take too long to move up on the SERP on the basis of user click-through rate.” Instead, Google used the occurrence of competing verticals to automatically boost its own properties above those rivals.

The staff report also documents a simultaneous strategy to demote comparison-shopping competitors. Through an algorithm launched in 2007, Google demoted all comparison-shopping sites beyond the top two positions on the SERP. Google’s own vertical properties were exempt from these same demotion algorithms—even when they met the similarity criteria. In fact, Google’s web spam team originally refused to include Froogle in search results, noting “[o]ur algorithms specifically look for pages like these to either demote or remove from the index.”

These changes had measurable effects. Google Product Search went from seventh in page views in July 2007 to number one by July 2008; its leadership acknowledged that “[t]he majority of that growth has been driven through product search universal.” Induced traffic losses and reduced competition, combined with demotions, led rivals such as NexTag to curtail investment in shopping comparison and, after Google launched Flight Search, to halt development of a competitive travel service entirely.

Scraping and Content Appropriation

The staff report’s strongest recommendation for a complaint concerned Google’s scraping of rival vertical content, particularly for local search. Google acknowledged internally that review content was “critical to winning in local search” but that it had an “unhealthy dependency” on Yelp. Internal emails show concern that Yelp could “become [a] competing local search platform[],” and Google executives made a failed attempt to acquire Yelp.

Two employees testifying for the report tell the story. After Yelp discontinued its review feed and requested removal of its content from Google Local, Google initially refused—claiming it was technically infeasible to remove the content without also blacklisting Yelp from all web search results. Mayer later admitted under oath that this claim was false. Google then introduced a policy forcing any property that refused to license its reviews to Google Places to be removed from Google’s search results entirely. Yelp, CitySearch, and TripAdvisor were all told they could only opt out of Places if they were fully removed from the search engine.

Google subsequently built its own reviews product from the scraped content, aggregated from third-party sites without attribution, and “unblacklisted” Yelp once its UI was live. Staff concluded that, by mid-2011, Google had collected enough reviews through bootstrapping and no longer needed to display third-party content, particularly while under investigation for the practice.

API Restrictions on Cross-Platform Management

A third area of concern involved Google’s AdWords API terms and conditions, which prohibited third-party tools from copying campaign data into other search networks or co-mingling AdWords data with campaign data from competing engines. These “restrictive conditions” prevented tool developers and agencies from creating a unified interface for multi-platform advertising management. Notably, the restrictions did not apply to advertisers themselves—so large players like Amazon and eBay had developed their own in-house multi-homing tools.

Google’s internal documents suggest the company understood the trade-offs explicitly. A 2007 product manager email—endorsed by director Richard Holden—stated: “If we offer cross-network SEM in [Europe], we will give a significant boost to our competitors... For this reason, [Microsoft] and Yahoo still have a fraction of the advertisers that we have in [Europe].” The email also acknowledged that the transactional overhead of managing multiple networks outweighed the “small amount of additional traffic” for most advertisers, effectively chaining them to Google’s AdWords auction.

Holden’s December 2008 evaluation for Susan Wojcicki (then SVP of Ad Products) concluded that removing the restrictions would “open up the market,” improve efficiency, and could be offset by Google competing with its own best-in-class SEM tool. However, the effort to relax restrictions was rejected by Larry Page, according to a January 2010 meeting record. One internal email describing that meeting notes “we didn’t take notes for obvious reasons,” suggesting awareness of the sensitivity of the subject. Google’s subsequent investment in DART Search—its cross-network tool acquired via DoubleClick—was explicitly discussed as constrained by the restrictive conditions, and internal documents link the desire to improve DART with the need to relax the API Terms & Conditions.

Syndication Agreements and Search Intermediation

The final major area concerned Google’s syndication deals. Staff reviewed Google’s AdSense for Search (AFS) agreements, which fall into two categories: negotiated Google Service Agreements (GSAs) held by roughly the top 10 partners (nearly 80% of query volume in 2011), and standard online contracts that are non-negotiable but cover only a tiny revenue fraction. All GSAs included terms effectively mandating exclusivity or “preferred placement” for Google—meaning non-Google search and ads were prohibited on designated pages, or Google ads had to be displayed in an unbroken block in the most prominent position, at a number at least matching any competitor’s.

As of 2008, Google shifted its favored arrangement from formal exclusivity to “preferred placement.” Staff found that a minority of the largest publishers objected to these terms; those that did—eBay, NexTag, business.com, and Amazon—described meaningful adverse consequences. eBay, which represented 27% of U.S. syndicated search queries, was contractually required to give AdSense ads preferential treatment. NexTag succeeded in removing an explicit exclusivity clause in 2010 but viewed the revised terms as “essentially the same thing as exclusivity.” Amazon wanted a longer-term deal but was offered only a one-year extension, conditioned on sending Google 90% of its search queries; Amazon refused the explicit quota but has been de facto hitting it anyway. IAC also described being contractually bound to Google on a per-property basis, and its local search unit, CityGrid, was prevented from testing competing syndication networks.

Publisher responses generally confirmed that Bing monetizes at significantly lower rates—Amazon reported Bing monetizing at roughly half of Google’s rate—making Microsoft’s network a hard sell even absent contractual ties. Between Q1 2009 and Q1 2010, Google systematically cut its revenue-sharing (traffic acquisition cost, or TAC) payments to AFS partners from 80.4% to 74%, yet no publisher considered the reduction decisive enough to switch. Internal Google guidance described the goal starkly: “Our general philosophy with renewals has been to reduce TAC across the board.”

For Microsoft, even a small query-volume gain was framed as highly meaningful, and the staff report’s investigations found that the syndication exclusivity provisions “make it less likely that small local competitors like IAC’s nascent offering can viably emerge.” In sum, the report’s evidence suggests that scale in search—whether in query logs, click data, or ad inventory—functions as a self-reinforcing economic moat, one that Google protected through a combination of product placement, content policies, and restrictive commercial terms.

Search, Monopoly and the Law: Dissecting the FTC Staff Report

The FTC staff's investigation into Google's search practices rests on a straightforward reading of Section 2 of the Sherman Act, which prohibits monopolization and attempted monopolization. To prove monopolization, the FTC must show both that Google possesses monopoly power in a relevant market and that it willfully acquired or maintained that power through exclusionary conduct rather than through superior products or business acumen. An attempted monopolization claim would additionally require showing predatory conduct, specific intent to monopolize, and a dangerous probability of success.

The staff report organizes its analysis around three questions: whether Google holds monopoly power, whether it engaged in exclusionary conduct, and whether that conduct violates the law. The report's conclusions are notably mixed. It finds that Google did violate Section 2 in several areas, including its scraping of rival content, its restrictive API terms, and its exclusive syndication agreements. But on the question of Google's preferential treatment of its own vertical properties in search results, the report declines to recommend action.

Defining the Markets Where Google Plays

Under antitrust law, a relevant market consists of a grouping of sales where a hypothetical monopolist could profitably raise prices significantly above the competitive level. Courts typically examine practical indicia such as industry recognition, product characteristics, distinct customers, and price sensitivity. The FTC staff identified three such markets in which Google operates.

The first is horizontal search. Vertical search engines do not offer a viable substitute for horizontal search, because they are not equipped to expand into general-purpose search, and they lack the comprehensive coverage that drives user demand. Eric Schmidt himself acknowledged this dynamic, testifying that a strong general product creates brand and demand, but vertical properties generally depend on horizontal engines for traffic, since users begin their searches at Google, Bing, or Yahoo. Google's own internal monitoring focuses on Bing and Yahoo, not vertical properties. In the United States, Google's share of general search is estimated at 66.7% by ComScore, with an additional 4.6% coming through its syndication to Ask.com and AOL. Yahoo holds 15% and Bing 14%, leaving Google well above the threshold routinely treated as sufficient for monopoly power.

The second market is search advertising. The staff concluded that search ads are distinct from display, contextual, behavioral, and social media ads, based on their scale, targetability, and the user's intent at the moment of display. Ads appear when a user is actively expressing interest, and they convert at significantly higher rates. Numerous advertisers confirmed they would not shift spend away from search ads in response to a small but significant price increase, with high-profile experiments supporting this. Chevrolet lost 30% of its clicks when it suspended search ads in favor of display alone. Google's own executives testified that search advertising is the most effective advertising tool with the best return on investment. Industry trackers place Google's AdWords share at between 76% and 80%, with the Bing-Yahoo partnership trailing at 12-16%.

The third market covers syndicated search and search advertising, or search intermediation. Here, horizontal search providers sell their services to other websites, sharing revenue from the ads those sites return. Publishers consistently reported that monetization through search ad syndication outperforms display advertising and other content, and none said they would leave the platform for alternatives in response to a modest price increase. Google's systematic reductions in traffic acquisition costs serve as a natural experiment confirming its dominance, and ComScore data shows Google's AdSense holds approximately 75% of this market, versus 22% for Microsoft and Yahoo combined.

Barriers That Keep Rivals Out

Entry into these markets is not easy. Developing a competitive search or search advertising platform requires substantial investments in specialized knowledge, technology, and infrastructure, as well as time to achieve meaningful scale. Google alone spent over $5 billion on R&D in 2011, while Microsoft invested more than $4.5 billion in 2010 on the algorithms and infrastructure for Bing.

The markets exhibit significant scale effects. Greater usage improves algorithmic quality and ad serving accuracy, which attracts more advertisers and users, creating a virtuous cycle that is difficult for new entrants to disrupt. According to Microsoft, the scale problem is the greatest barrier, and the company conceded it is losing $2 billion per year in its effort to compete. Reputation, brand loyalty, and exclusive agreements with high-volume syndication partners further entrench Google's position.

When Design Choices Cross the Line

Competition law distinguishes between conduct that excludes rivals on efficiency grounds and conduct that impairs rivals without furthering competition on the merits. Applying this standard, the staff's findings are uneven.

On Google's promotion of its own vertical properties within its search results, the staff concluded that, despite evidence of anticompetitive effects, bringing a case would be difficult. Google had begun moving toward universal search before it faced competitive threats from vertical properties, and the company argued the changes improved user experience. Internal documents show that Google feared losing lucrative queries to aggregation sites such as UK Finance comparison engines, and its preferential placement of its own properties did drive significant traffic away from competing verticals. Google's Product Search, which the product team could not even keep indexed in web results, became the most-visited comparison shopping site on Google after inclusion as a Universal Search result.

Yet the staff found Google's justifications compelling. Google argued its results provided consumers with better answers than blue links to other shopping sites. The company said technical and latency constraints made it infeasible to incorporate third-party data, though the staff noted Bing successfully serves Kayak results in its flight vertical. Google also pointed to an "apples and oranges" problem, asserting that comparing its universal results against web search results ranked on different criteria would be impossible. Most troubling to staff was that Google demoted vertical competitors that were, by the company's own admission, effectively identical substitutes.

Still, the report acknowledged that product design is an area where courts hesitate to second-guess a dominant firm's decisions, and Google's actions present a complex picture of a company attempting to maintain market share by improving user experience. The determination that this conduct was anticompetitive would require extensive balancing, and staff concluded that courts have been unwilling to perform that analysis under Section 2.

The scraping of rivals' content falls on the opposite side of the line. Google used its monopoly position to take content from sites like Yelp, TripAdvisor, CitySearch, and Amazon to upgrade its own offerings, harming competitors' ability to compete on innovation. This constitutes a conditional refusal to deal, since Google threatened to cease cooperating with companies that declined to let it copy their data. The report draws parallels to Aspen Skiing and subsequent appellate decisions holding that termination of a voluntary and profitable course of dealing to achieve an anticompetitive end violates Section 2.

Google's asserted technical difficulties are undercut by the record. After Yelp sent a formal cease-and-desist letter, Google removed Yelp's content almost immediately, and similar swift compliance followed when Amazon objected to its product reviews appearing in Google Product Search.

Contractual Restraints in APIs and Syndication

The AdWords API itself is a procompetitive development, but its usage restrictions are not. The API agreement limits advertisers' ability to use their own data and prevents third parties from offering campaign management tools that work across multiple search networks, hindering cross-network optimization. Google itself has been constrained by its own restrictions, holding back on improvements to its DART Search tool despite internal estimates that such functionality would benefit Google and advertisers.

These restrictions stopped a nascent market for cross-network management tools, forced search engine marketers to remove campaign cloning features, and increased transaction costs for advertisers not large enough to build their own infrastructure. Notably, because Google would not be required to grant rivals access to its API, the solution does not raise compelled-dealing concerns — Google could simply remove the restrictive conditions while keeping the API exclusive.

Google's sole justification is concern about misaligned incentives, fearing that agencies would adopt a "lowest common denominator" approach that degrades AdWords performance. The evidence indicates this rationale is pretextual. Internal emails reflect that Brin and Page championed a protectionist strategy to prevent third parties from developing consolidated management tools for both Google and its competition. A 2004 document discussed how to prevent a new MSN ad network from benefiting from a shared cross-network platform, considering requirements that applications include Google-centric features or outright disallowing cross-network compatibility.

The investigation into exclusive syndication agreements produces equally strong findings. Google's AdSense agreements frequently include exclusivity provisions that foreclose a substantial portion of the market for search intermediation. These agreements lock up major publishers, preventing them from credibly threatening to shift traffic to Bing or Yahoo to secure better terms from Google. Notably, Google has been reducing revenue shares to publishers without significant resistance, and internal documents show the company pursued an AOL renewal that it knew would be a "low/no profit partnership" precisely to prevent Microsoft from gaining scale.

The staff acknowledged uncertainty about whether ComScore or Microsoft's own data more accurately measures market foreclosure, but the qualitative picture is sufficiently concerning. While removing exclusivity might initially prompt only modest shifts in traffic, a Competitive Effects analysis looks further. Publishers would gain the opportunity to test alternatives to Google's AdSense program, which could significantly reshape market dynamics for new entrants.

Google's business justifications fare poorly against the evidence. The industry practice of exclusive guarantees dates from when a non-exclusive agreement like the 2008 proposed deal Google attempted with Yahoo likewise undermines the argument. The user confusion rationale, premised on concerns that inferior competitor ads would harm Google's brand, applies equally to situations where Google actively sought arrangements where partner sites would label whether advertising came from Google's network.

The overall evidence shows a search and search advertising market exhibiting classic signs of a monopoly protected by high entry barriers, where the dominant firm exercises pricing power. Under a complete rule-of-reason analysis, the weight of the evidence supports finding Google liable for imposing exclusive restrictions that prevent rivals from achieving the scale needed to compete effectively.

Remedies and Litigation Risk

If the FTC had pressed forward with the investigation, staff at the time identified several potential remedies. None of them involved technical fixes to Google's search infrastructure. The first category concerned scraping. Google could be required to offer an opt-out that removes snippets of content from its vertical properties while retaining web search results or Universal Search results on the main SERP. Alternatively, Google could be limited in how it uses indexed content—using it only to return properties in search results, not to inform its own product or local rankings—unless given explicit permission.

The second category targeted API restrictions. Staff noted that eliminating problematic contractual restrictions would require no technical fixes, since SEMs reported that cross-compatibility technology already exists and would quickly flourish if not for Google's contractual constraints. Remedies for exclusive and restrictive syndication agreements would enjoin Google from entering exclusive search syndication deals and loosen restrictions on AdSense partners using rival search ads.

But from a litigation perspective, the theory of harm had problems that the Bureau of Economics (BE) memo later highlighted in depth. Google did not charge customers, and they were not locked in. Universal Search produced substantial user benefits. Google's aggregation of content added value for its customers. The largest advertisers used both AdWords and Microsoft's AdCenter. The most efficient distribution channel through which Bing could have gained scale was its own site, and Microsoft had resources to buy distribution if it saw value. Most publishers used AdSense.

The staff's final conclusion nonetheless framed Google's conduct as unlawfully maintaining a monopoly over general search, search advertising, and search syndication under Section 2 of the Sherman Act, or otherwise engaging in unfair competition under Section 5, based on three theories: scraping rival vertical websites to improve Google's own offerings, entering exclusive agreements with web publishers that prevented them from displaying competing search results or ads, and maintaining contractual restrictions that inhibited cross-platform ad campaign management. The staff recommended the Commission issue a complaint. The memo was submitted by Barbara R. Blank and approved by Geoffrey M. Green and Malanie Sabo.

The FTC ended up closing the investigation without action.

The BE Memo's Pushback

The Bureau of Economics memos from August 8, 2012 offer a window into what actually happened. The anticompetitive investigation began in June 2011. Staff presented theories and evidence in February 2012. This memo, from economists Christopher Adams and John Yun, was a final recommendation—and it concluded the investigation ought to be closed.

Four harm theories were considered: preferencing Google's own properties in search results; exclusive agreements with publishers and vendors that deprived rivals of users and advertisers; restrictions on porting advertiser data to rival platforms; and misappropriating content from Yelp and TripAdvisor. The memo's guiding principle was that the analysis had to move "beyond collecting complaints and antidotes" [sic] from competitors who were negatively impacted by Google's practices.

The BE's market definition showed Google with a "significant" share—65% of paid clicks and 53% of ad impressions among the top five U.S. search engines. However, the memo argued market power was mitigated by the fact that 80% of users used a search engine other than Google, and empirical evidence was consistent with search and non-search ads being substitutes. Google's internal analysis also treated vertical search as competitive pressure.

For the preferencing theory—where Google blended its own content with standard "blue links" and demoted competing vertical sites—the BE claimed Google accounted for only 10% to 20% of traffic to vertical rivals, making the harm "very small and not statistically significant." Universal Search was characterized as a procompetitive response to vertical site pressure, and an improvement for users.

On the exclusive agreements theory, the memo noted that direct access to a search engine site—not third-party distribution deals—remains the most efficient distribution channel, where Google's conduct did not impede. Moreover, search toolbars and default browser status were not "exclusive" in any meaningful sense, because users could easily switch. This asserted ease of switching is remarkably out of step with how search defaults actually functioned then and now. Google's actual blocking of default search engine changes on mobile was already a documented practice. If switching were truly effortless, companies would not have paid on the order of tens of billions of dollars to secure default status in browsers and on mobile platforms like iOS. A key objection at the time, now vindicated by events, is that the entire incentive structure of the search ecosystem—contracts for default placements, vendor lock-in efforts, and platform pressure such as Google's influence over Samsung—contradicts the assumption of fully informed users easily migrating. Consumers are not all rational, fully informed agents who switch at will; there's a reason defaults command enormous value.

After Microsoft's alliance with Yahoo was announced, the memo noted Microsoft and Yahoo combined had a higher search share in syndication than Google, and their query volume grew faster than Google's; users had gone from 2.5% query share to adding one point per year. Microsoft also enjoyed multi-year partnerships with Facebook and was the default search engine on devices like the Xbox and Kindle Fire. Their combined market share with Yahoo had been steady at 30%. In December 2011, Microsoft had query volume equivalent to Google's two years earlier—leading the BE to infer that Microsoft was not below a problematic scale threshold. This missed the forward-looking dynamics that context makes obvious. In fast-moving tech markets, equivalent query volume from two years prior does not imply competitive position. A rival that holds a market share below some threshold now may not stay there, but one keeping up only with the absolute volume a leader posted previously is plainly vulnerable to scale-driven moats. The memo correctly asserted Microsoft had not been foreclosed from search syndication—Marvel, Yahoo, AOL, InfoSpace, and others used its ad platform—but competition concerns were nonetheless premised on exclusive agreements; the memo failed to find the factual support for the broader claim.

For restrictions on AdWords API portability, the BE argued the introduction of the API, with its co-mingling restriction on data from other engines, made users and Google better off, and that rivals' costs were unaffected. Advertisers accounting for the "overwhelming majority" of search ad spend used both Google and Microsoft, and evidence from SEMs suggested the policy had negligible impact on Microsoft ad spend. The two theoretical theories of harm pertaining to AdWords API data porting restrictions were rejected. The existing exception for "beta" SEMs that historically used Marin, Kenshoo, or SearchIgnite did not reflect that the market itself had responded productively to any alleged restrictions, per the BE's interpretation.

On scraping, the BE had substantive concerns about Yelp and TripAdvisor content but believed the antitrust case required strong evidence that misappropriated content shifted users from Yelp and TripAdvisor to Google, or that it reduced their innovation incentives. No such evidence existed in the record, staff concluded.

Conclusion

The BE memo also unsteadily dismissed the vertical facts. Since then, Google's market share has remained well above 90% without dipping below the levels cited as stable at the time the BE made its recommendations. The economies of scale in search continued to accrue, and the experts who authored the policy exemptions embraced a model of behavior that real market participants have systematically disproven or challenged. The case was closed not because Google's market position was safe from scrutiny, but because the analytical framework—applied by the FTC's economist staff—could not see the competitive harms that a careful reading of the market dynamics makes visible.

Monopoly power: what the evidence actually shows

The first question the FTC had to answer was whether Google holds monopoly power in a relevant antitrust market under Section 2 of the Sherman Act. The Bureau of Economics memo leaned heavily on the idea that online search is just another form of advertising, and that the relevant competition happens between platforms and advertisers trying to reach the same users.

The argument is that Google's market power depends on its share of internet users, not its share of search queries. If advertisers can reach a given user on Yahoo, Bing, or Facebook, then Google's position is weaker than its raw search share would suggest. The memo notes that Google's share of paid search clicks among the top five general search engines grew from 55% in September 2008 to 65% in February 2012, and argues that this simply reflects Google's ability to deliver what advertisers want: eyeballs.

The multi-homing evidence is central here. Roughly 80% of users visit at least one non-Google search platform in a given month, suggesting that advertisers can reach the same users elsewhere, even if via a different query on Yahoo or Bing. But this reasoning is more fragile than it appears. If a user defaults to Google and only occasionally lands on another engine, that does not meaningfully expand an advertiser's reach. The underlying data comes from comScore, which reports that 20% of users only use Google, 15% never use Google, and 65% use both. ComScore's accuracy on search share is widely questioned; for example, the firm shows Google with an implausibly low 67% share today, while analysis of real traffic to sites shows Google driving roughly 85% of clicks.

The memo also claims that firm-level spending on search ads and display ads is negatively correlated, so the two cannot simply be dismissed as unrelated. It stops short of declaring them the same product market, but argues that substitution is real. Vertical search engines are treated as differentiated competitors, with the analogy of a supermarket competing with a convenience store: for a query like "Nikon 5100," Amazon offers a competing but differentiated product to Google's results. The bottom line is that while Google leads in search, that lead is mitigated by multi-homing, display substitution, and vertical competition.

The preferencing theory

The first theory of harm is that Google's blending of vertical content — shopping comparison results, local business listings — into its main results page disadvantages competing vertical sites like Nextag, eBay, Yelp, and TripAdvisor. The blend has two effects: it pushes those sites down the page, and it changes Google's incentives to display competing vertical links at all. Three empirical questions matter:

• How much of vertical sites' traffic does Google account for?
• How do blends affect click-through rates to vertical competitors?
• Do blends improve the value consumers get from search results?

The memo's data shows Google search drives 10% of traffic to shopping comparison sites and 17.5% to local business search sites. Internal Google data on the effects of blends shows significant drops in clicks to other shopping comparison sites: a site with a 9% pre-blend click-through rate would see it fall to roughly 5.3% when a blend appeared. Local blends have a smaller effect, reducing a 6% click-through rate to about 5.4%.

But the memo is careful to note that a decline in click-through rates when a blend appears is not the same as a decline in overall traffic from Google. If blends are a quality improvement, they increase query volume on Google, which benefits all sites. The memo then attempts to infer consumer value from click behavior: if clicks on ads decrease when blends appear, users must prefer the content, so blends must be increasing consumer welfare.

That inference is methodologically suspect. It only works if blended content is presented with the same visual weight, position, and prominence as ordinary algorithmic results, which is not the case. Blends occupy a distinct visual format and position, so comparing click patterns between them and standard blue links is an apples-to-oranges exercise. The conclusion may be right — blends probably do improve consumer value — but this particular argument does not establish it.

The documentary record and the verdict

As a matter of history, the memo notes that general search engines have blended vertical content since the 1990s, and all major engines (Google, Yahoo, Bing) use blends today. This is relevant context for evaluating whether Google's behavior was exclusionary or merely competitive.

On the preferencing theory specifically, the FTC team's conclusion is that Google is not a significant enough source of traffic for its blending practices to foreclose vertical rivals. That conclusion rests on a platform traffic model that the lawyers' memos elsewhere treat as flawed — the model seems to underestimate Google's role in referral traffic, which would undercut the claim that competitors could not be meaningfully harmed.

Distribution deals and the scale question

The FTC’s second major theory held that Google used exclusionary distribution agreements to deny Microsoft the scale it needed to compete effectively in search. The underlying assumption was that search quality improves with query volume, creating a feedback loop that an incumbent could exploit to lock in its advantage. The Bureau of Economics (BE) staff pushed back on this theory, and their reasoning is worth examining in detail.

The exclusionary-agreement argument

BE staff acknowledged that exclusionary agreements merit scrutiny when they “materially reduce consumer choice and substantially impair the opportunities of rivals.” But they argued that Google’s deals did not meet that bar. On desktop, they noted that 73% of search traffic comes through direct navigation, a channel equally accessible to every search engine. In their view, Google could not impair rival access to the most important distribution channel even if it wanted to.

This model has a basic problem: if direct navigation were truly the dominant and most efficient channel, companies would not spend enormous sums to secure default placements. The market behavior suggests the default position matters a great deal, and that users often continue with whatever search engine comes preconfigured rather than actively choosing one. Defaults influence future direct-navigation habits, which the BE staff model does not capture.

The staff also pointed out that Microsoft was the default search engine on Internet Explorer and on 70% of PCs sold. That is accurate but does not address the asymmetry that emerged later as browser and OS defaults shifted toward Google.

Syndication and premium placement

On syndication agreements, BE staff argued that Google’s premium-placement requirements were negotiable and that subjecting an entire publisher site to a single bundled deal intensified ex ante competition for the contract, lowering publishers’ costs. Eliminating bundled discounts or exclusivity, they claimed, would result in higher prices to publishers.

That analysis contradicts observed practice. Publishers have repeatedly complained about the terms of these deals, and the suggestion that site-wide exclusivity benefits publishers by creating more intense competition is not supported by how these negotiations actually play out. The staff concluded that any scale advantage Google obtained was the result of “competition on the merits,” equivalent to winning traffic by building a better product at a lower price. That framing ignores the feedback loop at the heart of the complaint — that scale itself improves the product, making it difficult for a smaller rival to catch up regardless of merit.

Market share and the mobile blind spot

BE staff cited market share data to argue that Microsoft and Yahoo were not being excluded. They noted that combined syndication shares for Microsoft and Yahoo were higher than their combined general search market share. At the time, the numbers showed Google at 44% including AOL and Ask, with Microsoft and Yahoo combining for roughly 56% in syndication. Those figures were already doubtful when written and became less directionally accurate over time as Bing and Yahoo lost ground.

The mobile section is where the analysis most clearly misses the trajectory of the market. Staff noted that mobile search accounted for only 8% of queries and an even smaller share of search ad revenue, treating it as a minor issue. That characterization was obviously wrong at the time. Google had already shifted to a “mobile first” strategy across its products, making changes that sometimes degraded the desktop experience to improve mobile. This was widely understood inside the industry as the only reasonable long-term direction. Economists studying the market who interviewed people at Google and other tech companies should have seen it.

The staff also claimed that switching costs on mobile were trivial — “a few taps” and downloading other search apps in seconds. The sums companies were willing to pay for mobile default positions strongly suggest otherwise. If switching were truly frictionless, those default slots would be worth little.

Evidence of exclusion in practice

On the question of whether rivals were actually being excluded, BE staff argued that Microsoft and Yahoo’s steady 30% combined share over four years showed no sign of exclusion. That stability is not evidence of health; the writing was already on the wall for both Bing and Yahoo. The staff cited comScore data suggesting Microsoft query volume grew 134% while Google grew 54% since the Microsoft-Yahoo alliance was announced. Those metrics were inaccurate and misleading at best — they counted unique users in a month, treating someone who searched once as equivalent to a daily user.

The staff acknowledged that Microsoft could not, in a single meeting with Susan Athey, produce data showing how its cost curve changed as click data changed. They treated this as grounds to doubt that Microsoft was below any scale threshold. That leap is weak on two counts: the concept of a discrete “threshold point” misrepresents how scale effects work, and the inability to answer a question in one meeting is thin evidence.

The staff also cited Microsoft’s own press releases — including Steve Ballmer’s statement that the Yahoo deal would “provide the scale we need” — to argue that Microsoft was not actually scale-constrained. Using promotional material as evidence against a well-supported claim is unusual. Engineers who worked on search at both Google and Bing consistently said that more volume and more data were major advantages; a corporate press release is not a credible counterweight to that.

Feedback effects and raising rivals’ costs

BE staff considered several feedback mechanisms that could make distribution agreements anticompetitive even if no single agreement was exclusionary. They examined a scale effect (lower cost per unit of quality or ad matching with more data), an indirect network effect (more advertisers attracting more users), a congestion effect, and a cash flow effect.

The scale effect was dismissed, with reasoning that was simply wrong. The indirect network effect was found to have weak evidence, and the staff noted that low ad click-through rates showed most consumers do not like ads anyway — a non sequitur that does not bear on whether more advertisers improve matching quality. They also argued that a greater number of advertisers leads to congestion, reducing platform value for advertisers. There is a narrow sense in which that is true, but platforms with few advertisers are far less attractive to both advertisers and users, as the subsequent history of Bing demonstrated.

The cash flow effect was dismissed because Microsoft was not cash-flow constrained — a strange argument given that Microsoft soon sharply cut Bing investment because returns did not justify the cost. Economists should recognize that marginal cost and marginal revenue matter even for well-funded firms.

Other claims

The authors of the BE memo also argued that removing API interoperability restrictions might cause long-term harm by shifting incentives and reducing innovation, which must be weighed against short-term benefits. That argument mirrors later claims that banning non-compete clauses would reduce firm incentives to innovate.

The memo relied on Microsoft adCenter advertising materials claiming easy data import from AdWords to argue that data portability was not a real issue, calling these advertisements “more credible than” other evidence. Treating marketing copy as a reliable source — and elevating it above technical testimony — was a remarkable evidentiary choice. The authors also did not believe Google Search and Google Local were complements, or that displaying Yelp and TripAdvisor data directly in search results harmed those sites. They said the burden of proof to show such harm would be “extremely difficult.”

The BE memo’s analysis of exclusionary distribution deals rests on a series of assumptions that do not survive contact with how the search market actually worked: that defaults do not matter much, that mobile was a sideshow, that switching costs were trivial, and that marketing materials are reliable evidence. Each assumption pointed in the same direction — toward a finding of no competitive harm — and each was contradicted by the behavior of the companies involved.

Inside the FTC’s Google documents: Divergent views and missed calls

After the primary Bureau of Economics (BE) memo, the remaining FTC documents paint a more fragmented picture, with individual staffers and directors weighing in on specific aspects of the investigation. High-level summaries of these memos reveal a mix of narrow legal arguments, conflicting data points, and internal disagreement about whether any enforcement action was warranted at all.

Advertising practices and consumer deception

Laura M. Sullivan of the Division of Advertising Practices authored a memo taking a fairly narrow line on whether Google’s integration of paid results into organic listings constituted deception. The key conclusion: “we continue to believe that Google has not deceived consumers by integrating its own specialized search results into its organic results.” As a result, Sullivan recommended against further action on that front.

Sullivan did offer practical suggestions, noting that “based on what we have observed of these new paid search results [referring to Local Search, etc.], we believe Google can strengthen the prominence and clarity of its disclosure.” In hindsight, the opposite has occurred, rendering the advice moot.

The memo also revisited a 2002 FTC staff letter concerning disclosure of paid search results. Sullivan argued that updating that guidance was warranted, citing studies demonstrating that the methods used by Google, Bing, and Yahoo! to mark paid listings were insufficient. Internal Google research supported this concern; one June 2010 note from a senior member of Google’s in-house research group admitted:

“I don't think the research is inconclusive at all - there's definitely a (large) group of users who don't distinguish between sponsored and organic results. If we ask these users why they think the top results are sometimes displayed with a different background color, they will come up with an explanation that can range from 'because they are more relevant' to 'I have no idea' to 'because Google is sponsoring them.'”

On the subject of fraudulent ads for fake weight-loss and mortgage-relief products, Sullivan’s memo argued against pursuing Google. The rationale: “there is no indication so far that Google has played any role in developing or creating the search ads we are investigating,” the company had taken some measures to block them, and Google could claim CDA immunity. Thus, further investigation into this area was not worthwhile.

A separate memo from the same author examined whether Google’s use of consumer data in conjunction with its search advertising business was unfair. The conclusion was that it was not; consumers should reasonably expect their data to improve search queries.

Economic analyses and conflicting data

Ken Heyer, then a Director of the Bureau of Economics, contributed two memos. The first generally aligned with the primary BE memo, but pushed for caution before any complaint was filed. Heyer argued that the agency should have a remedy that seems “quite likely to do more good than harm” before “even considering serious filing a Complaint.”

On distribution, Heyer agreed that mobile was not yet an important channel and that Microsoft had strong desktop distribution via Internet Explorer’s default status on 70% of PCs sold. His view on API restrictions was mixed. Regarding mobile specifically, Heyer suggested gathering data on how much Google paid for default status, reasoning that large payments would indicate significant competitive value, and also wanted to measure how often users switched from the default. He cited mobile as only 8% of the market — a figure that was likely already outdated when written in late 2012, when mobile likely accounted for 20% or more of queries. Heyer’s analysis of vertical search sites matched the data analysis in the BE memo.

Heyer’s second memo was more emphatic, recommending that the FTC take no action whatsoever—not just avoiding litigation, but also declining a consent decree. A follow-up memo from BC staff, authored by Barbara R. Blank and colleagues, took the opposite stance, recommending that staff negotiate a consent order with Google specifically on mobile distribution.

The mobile distribution evidence

The BC staff memo presented a strong case that Google held exclusive agreements with the four major U.S. wireless carriers and with Apple to pre-install Google Search as the default. The Apple agreement explicitly required exclusivity, and Google Search was the default on 86% of devices. The recommended remedy was a consent agreement eliminating these exclusive deals.

These memos contain a striking data contradiction. The BE memo claimed mobile was roughly 8% of queries, but the BC staff memo cites Google’s own documents showing mobile was 9.5% of queries in 2010 and 17.3% in 2011. This rapid growth indicated that the mobile distribution channel was meaningful and, as both Microsoft and Google internal documents acknowledged, mobile would likely surpass desktop in the near future.

Google’s internal communications further undermined the claim that carrier agreements did not mandate exclusivity. While carrier statements suggested otherwise, the contracts told a different story: Sprint and T-Mobile agreements appeared to mandate exclusivity; AT&T’s deal was de facto exclusive due to a tiered revenue-sharing arrangement; and Verizon’s agreement was explicitly exclusive. Chris Barton, a Google business development manager, spelled this out:

“So we know with 100% certainty due to contractual terms that: All Android phones on T-Mobile will come with Google as the only search engine out-of-the-box. All Android phones on Verizon will come with Google as the only search engine out-of-the-box. All Android phones on Sprint will come with Google as the only search engine out-of-the-box. I think this approach is really important otherwise Bing or Yahoo can come and steal away our Android search distribution at any time, thus removing the value of entering into contracts with them. Our philosophy is that we are paying revenue share”

Additionally, Andy Rubin, then head of Android, had laid out a plan to reduce carrier revenue share over time as Google’s search dominance grew, a plan that was subsequently executed. Internal notes also suggested carriers would be unlikely to switch even if agreements were non-exclusive due to Google’s superior monetization or the risk of negative publicity. Rubin’s closing remark after finalizing the Verizon deal was telling: “[i]f we can pull this off ... we will own the US market.”

Legal and economic conclusions

Willard K. Tom, the FTC’s General Counsel, offered a measured final assessment. His memo concluded: “In sum, this may be a good case. But it would be a novel one, and as in all such cases, the Commission should think through carefully what it means.”

Howard Shelanski, a Director in the Bureau of Economics, largely supported the BE memo and Heyer’s analysis, with one notable exception: on the question of scraping, he sided with the BC staff’s position.

The spectrum of opinions across these documents is broad — from recommending no action, to pursuing a narrow disclosure fix, to seeking a consent decree on mobile. The inconsistency in key figures, such as the 8% versus 17% mobile share, raises serious questions about the reliability of the economic analysis that underpinned the decision to close the investigation.