Case Study 1: The Review Economy

Yelp, Google, and what it means to have your most important marketing asset owned by somebody else

Background

Sometime in the last twenty years, the most consequential piece of marketing collateral a restaurant has stopped being something the restaurant produced.

It is not the menu, the website, the sign, or the photograph in the window. It is a number between 1.0 and 5.0, computed by a private company from text written by strangers, displayed next to your name on a screen you do not control, in a ranked list you did not enter voluntarily. You cannot edit it. You cannot remove entries from it. You often cannot verify that the person who wrote an entry ever ate at your restaurant. And for a meaningful share of the people deciding where to have dinner tonight, that number is the entire evaluation.

Yelp, founded in 2004 and publicly traded since 2012, built its business on aggregating consumer reviews of local businesses, with restaurants as its most prominent category. Google subsequently folded reviews into its Business Profile product, placing a star average directly into search results and Maps — which, for most restaurants in most American markets, is now the higher-traffic surface of the two.

Neither company promises a restaurant anything. Both make money from the businesses they rank, which is the structural tension this case is about.

The operating issue

Three features of this arrangement create genuinely hard problems for an operator, and they are worth separating because the correct response to each is different.

1. The platform filters, and the filter is opaque

Yelp has long operated recommendation software that decides which submitted reviews are displayed prominently and which are pushed into a secondary "not currently recommended" list. Yelp describes this publicly as a measure against fake and solicited reviews, and on its own terms the goal is defensible: a platform whose ratings can be purchased or manufactured is worthless.

The consequence for an operator is that a genuine, enthusiastic review from a real guest may simply not appear. Reviews from accounts with little history, reviews clustered in time, and reviews that look solicited are exactly the ones a small restaurant is most likely to generate — because a small restaurant's happiest guests are often people who have never written a review before. The operator experiences this as the platform hiding good news, and there is no appeal.

2. The platform sells advertising to the businesses it ranks

This is the tension that produced litigation. Beginning around 2010, groups of small businesses alleged that Yelp manipulated the display of reviews — suppressing positive ones or surfacing negative ones — as leverage to sell advertising, and characterized the practice as extortion.

The claims were consolidated and litigated. In Levitt v. Yelp! Inc., the U.S. Court of Appeals for the Ninth Circuit affirmed dismissal in 2014, reasoning in substance that Yelp had no pre-existing obligation to display any particular review, and that aggressive sales conduct around a service a business had no legal right to receive did not amount to extortion. Yelp has consistently denied manipulating ratings for advertisers.

The legal outcome is worth stating precisely, because it is frequently misreported in both directions: the courts did not find that Yelp manipulated reviews, and they did not find that it did not. They found that the conduct alleged, even if true, was not the crime the plaintiffs named. Regulators examined the general question over the following years without bringing an enforcement action on the extortion theory.

3. The reviews themselves can be manufactured

The other half of the problem runs in the opposite direction: businesses buying favorable reviews and attacking competitors with unfavorable ones.

In September 2013, the New York Attorney General concluded a year-long undercover investigation known as Operation Clean Turf, which posed as a business seeking help suppressing negative reviews. The investigation identified search-optimization and reputation-management firms that were writing fake reviews — sometimes using overseas freelancers paid per review — and resulted in agreements with nineteen companies to cease the practice and pay penalties totaling several hundred thousand dollars.

More recently, the Federal Trade Commission finalized a rule in 2024 addressing fake and deceptive consumer reviews and testimonials, prohibiting practices including buying positive or negative reviews, undisclosed insider reviews, and certain forms of review suppression. The FTC's long-standing Endorsement Guides separately require disclosure of material connections between a business and a person endorsing it.

What it shows

The asset is real, it is enormously valuable, and you do not own it. A restaurant's star average functions economically like brand equity — it accumulates slowly, it decays if neglected, and it determines whether you make somebody's shortlist. But unlike brand equity, it sits on a private platform, is computed by a formula you cannot inspect, and can be moved by people who have never been inside your building.

The platform's incentives and yours are aligned only partly. A platform wants ratings that consumers trust, which is broadly good for a well-run restaurant. It also wants advertising revenue from the businesses it ranks, which creates a commercial relationship in which the ranked party has much less power. Neither of those facts is a scandal. Both are structural, and an operator should plan around them rather than be surprised by them.

Manufacturing reviews is both illegal-adjacent and strategically stupid. The enforcement record — Operation Clean Turf, the FTC's rule, the platforms' own removal policies — establishes real exposure. But the more decisive argument for a restaurant is arithmetic. The offset identity in §27.3 says that at a 4.6 average you need nine genuine five-star reviews to undo one one-star. A restaurant that buys reviews is spending money on the numerator of a fraction whose denominator is the actual experience in the dining room, and the actual experience is the thing that generates the one-stars.

And the practical exposure is worst when you are smallest. A restaurant with 40 reviews moves substantially on a single bad night. A restaurant with 500 barely moves. Every mechanism in this case study hurts a new independent more than an established one.

Outcome

Yelp remains a significant platform, though for most American restaurants Google's profile and review surface now carries more decision traffic. The litigation did not change the structure of the review economy; the regulatory activity has largely targeted the manufacture of reviews rather than the platforms' display of them. Restaurants continue to operate inside an arrangement where their most important public number is computed by somebody else.

What has changed is that the practices around it are now more clearly rule-bound. Solicitation is allowed on some platforms and discouraged on others. Gating is prohibited by policy. Incentivized reviews are prohibited by policy and, in several forms, by rule. Undisclosed material connections are a federal enforcement matter. The gray area has narrowed considerably, which is on balance good news for an operator who intends to do it properly.

The lesson

Manage the inputs, not the number. You cannot control the average, you cannot remove a review, and you cannot count on a platform to surface the good ones. What you can control:

1. Ask everyone, consistently, without incentive. The single largest determinant of your average is who gets asked, because unasked restaurants are reviewed disproportionately by the aggrieved.
2. Respond to the next reader. Three sentences, specific, never adversarial. The reviewer is not your audience.
3. Build volume early. The first hundred reviews are the fragile period, and volume is the only defense against a single bad night.
4. Fix the thing the review is about. The manager who touches a struggling table at minute twenty is doing more for the average than anyone with a marketing title.
5. Never manufacture. Beyond the enforcement exposure, purchased reviews are a bet that nobody will notice a gap between a 4.9 rating and a forty-minute ticket time, and the gap is what people write about.

And one structural conclusion that belongs in a business plan: because the review surface is rented, the owned channels in §27.5 matter more than they appear to. An email list of 2,400 addresses is the only version of your guest relationship that cannot be re-ranked, filtered, or repriced by somebody else.

Discussion questions

1. The Ninth Circuit's reasoning in Levitt turned on the fact that Yelp had no obligation to display any particular review. Is that the right legal answer? Is it the right commercial answer for a restaurant deciding how much to depend on the platform?
2. A platform that filters solicited reviews is protecting the integrity of its ratings. It is also systematically discarding the honest enthusiasm of first-time reviewers, who are exactly the guests a small restaurant generates. Can both of those be true? What should a platform do instead?
3. Operation Clean Turf targeted the firms writing fake reviews. Should it also have targeted the businesses that hired them? What would enforcement against a 40-seat restaurant accomplish that enforcement against a reputation-management firm does not?
4. Your competitor has a 4.8 average and you have a 4.5, and you are reasonably confident their food is worse. What are your options, in order of both cost and integrity? What would you actually do?
5. This chapter refuses to state what a star of rating is worth in revenue. Does the refusal make the chapter more useful or less? What would it take to establish the number honestly for one specific restaurant, and would it be worth the cost?