Skip to main content

5 Years Later: Google Raises its 2014 European Antitrust Settlement Proposal from the Dead

Last week we saw concerning reports that Google is preparing to propose a toothless settlement deal to the European Commission that is similar to —and even worse than — the settlements that the European Commission rejected 2013 to 2014. Yelp welcomes a meaningful remedy from Google that preserves competition.  Unfortunately, the latest proposal falls far short.

Back in 2014, a new Commission was set to take office in a matter of months.  Google attempted to exploit the change of administration by rushing through an ineffective remedy.  In 2013, not long after the U.S. Federal Trade Commission closed its investigation, former Google Chairman Eric Schmidt attempted to settle allegations of widespread search bias by proposing a settlement framework built around “rival links.” Google designed a settlement screen that showed links to third-party rivals (like TripAdvisor, Jameda, Yelp, etc.) while preserving the basic design of Google’s search engine results page: a “hardwired” answer box (or OneBox) which blends a helpful Google Map with local businesses and review information of questionable quality exclusively culled from Google.

The first and last iterations of the proposals are the only versions of the screens from the 2013-2014 negotiations in the public domain. Tests showed rivals would never receive over 2% of clicks (divided among three rivals)

The Commission sought feedback from the industry in the form of a “market test.”  Overwhelmingly, participants in the market test pointed out deep flaws in the settlement proposal. Iteration-after-iteration of Google’s concept would deliver no meaningful traffic to rivals.

Now in 2019, Google’s newly proposed remedies are just more of the same. How do we know?

The rise of services like Usability Hub and Mechanical Turk has created a Moore’s Law for antitrust enforcers, overcoming the information asymmetries once inherent in negotiating with large consumer internet platforms. Absent the emergence of such tools, Google would be holding all of the cards. Instead, anyone with some basic technical knowledge can effectively look “under Google’s hood” with rapid, inexpensive, and accurate estimations of user behavior. The fundamental insight is that search engine result pages are not unlike ballot screens. Just as an entire industry of political polling is dedicated to accurately predicting how people will vote in elections by prompting respondents with hypothetical scenarios, any regulator (or Jane Citizen, for that matter) can now accurately predict how people behave given a hypothetical search results page. A peer-reviewed paper based on such methodologies was published this quarter in the prestigious journal of Management Science.

By replicating these methodologies (i.e., hooking MTurk to UsabilityHub), we demonstrated some counterintuitive findings: Google could increase the amount of real estate a particular rival received on a Google search results page while significantly decreasing the rival’s average click-through rate. This user interface sleight of hand was one of the shocking findings that the Commission relied on to reject the early rival links proposals from Google.  

With input from the market, the Commission saw through Google’s tactics and did the right thing.  The Commission rejected Google’s settlement proposals and brought formal charges against Google for manipulating its shopping comparison search results. It has been reported that the Commission has since opened a probe into local search.  Fearing the same fate as shopping, Google is recycling its remedies proposal from 2013 in an attempt to avoid formal charges in local search.

Why is a fruitless negotiation over “rival links” relevant five years later?

Google is utilizing the same playbook. Except this time it’s worse. We know because the experiment screens are out in the wild, and we’ve already begun testing them.  

A summary of these results is viewable here.

The latest rival links proposal, which has been reported on by multiple media outlets, still receives a slightly improved but ultimately marginal percentage of clicks (around 5%), while Google’s capture of web traffic has increased by approximately 20% since the original proposal over five years ago. Seven in ten clicks go to Google (compared to about five in ten clicks from the ‘13-’14 proposals). These trends are consistent with other empirical evidence showing Google is eating more and more of the web in order to steer traffic to itself.

In all fairness, the primary screens provided by Google during the 2013-2014 period were desktop mockups and the experiments out in the wild are from smartphone screens. So while it is important to acknowledge such a distinction, it is also critical to note that smartphones’ tinier form factors enable Google to siphon even more traffic away from the web. Since 2014, there was also a critical inflection point where smartphones overtook desktop as the primary form of searching. Estimates suggest local search is half of all searching on smartphones. Thus, whether it is on desktop or mobile, the “rival links” proposal by Google is and always has been about distracting from the elephant in the room: why Google still refuses to create an interoperable, non-discriminatory Local Universal box which both answers users’ queries while sending traffic to the web.

Such a solution is not only technically trivial for Google to implement, it would ensure that candidates for European consumers’ local search results are pooled from the entire web, instead of from Google’s walled garden of inferior reviews.

So the gall of Google’s proposal is not only that it’s inferior to the offer already rejected, more than that, it’s not even a formal Commitments offer within the Commission’s rulebook. It’s an ad hoc change that Google could change or even abandon at will.

Yelp welcomes a meaningful and binding remedies proposal from Google that preserves competition and maximizes welfare for consumers, but the latest proposal from Google does neither.  It’s just another page out of the Google Playbook: propose a weak “rival links” remedy at the very end of a Commission’s administration.