One Star, Five to Nine Percent: What the Ratings-Revenue Literature Says About High-Consideration Local Services

How much do reviews affect revenue? Start with the one study that measured it

Most published claims about reviews and revenue are correlations dressed as causes. A business with more five-star reviews earns more, but higher-quality businesses also earn more and attract better reviews, so the raw association cannot separate the review's effect from the underlying quality it reflects. To measure the causal value of a rating point, you need a source of rating variation that is unrelated to true quality.

Michael Luca's study of Yelp supplied exactly that. Yelp displays a business's rating rounded to the nearest half-star, so two restaurants of nearly identical underlying quality can be pushed onto different sides of a rounding threshold and shown a visibly different rating for reasons that have nothing to do with the food. That rounding is a natural experiment. Comparing restaurants just above and just below each threshold isolates the effect of the displayed rating itself.

The result is the number in this article's title. A one-star increase in the displayed Yelp rating produced a 5 to 9 percent increase in revenue. It is, to date, the cleanest causal estimate of what a rating point is worth to an independent local business, and it is why the finding has become a reference point across local-commerce and platform-economics research.

Why the effect lands on independents, not chains

The most instructive part of Luca's finding is not the headline range but its boundary. The revenue effect was concentrated in independent restaurants. Chain-affiliated restaurants showed no measurable rating effect at all.

The proposed explanation is that consumers already hold a quality prior for a chain. A national brand has taught the buyer, through advertising and repeated experience, roughly what to expect before a single review is read, so the marginal review changes little. An independent has no such buffer. For an unfamiliar local business, the synthesized rating is close to the only quality signal the buyer has, so it carries the full weight of the decision. The study also found consumers respond more strongly to ratings backed by more visible and more numerous reviews, consistent with the rating mattering most precisely where it is the buyer's primary evidence.

For the businesses this publication serves, owner-operated and small-team local services rather than franchises of national brands, this boundary is the relevant half of the finding. It says the rating point is worth most to exactly the kind of firm that has no brand prior to fall back on. It does not say the 5 to 9 percent figure is theirs to claim.

The mechanism generalizes; the magnitude may not

There is a defensible reason to expect the underlying mechanism to extend beyond restaurants. The mechanism is not about food. It is about a buyer choosing an unfamiliar provider under uncertainty, using a reputation signal that a platform has already aggregated on their behalf. That structure describes how people now select a plumber, a dentist, or a gym as much as it describes how they pick a restaurant.

Adjacent research on provider selection points the same way. A large topic-modeling study of online health communities, covering 747 doctors and 105,032 reviews, found that narrative reviews measurably shifted which provider a patient chose, and that the content of those reviews predicted choice differentially. The synthesized reputation signal moved the decision, which is the same causal role the star rating plays in the Yelp result.

Two limits keep this from being a transfer of the number. First, that provider-choice study ran on a non-US platform, so its mechanism generalizes more safely than its magnitude. Second, and more fundamentally, a mechanism that operates in the same direction can operate at a very different strength. Booking a dental implant, hiring a contractor for a roof, and choosing a restaurant for Friday differ in price, risk, switching cost, and how many other signals the buyer consults. Any of those could make the revenue response to a rating point larger or smaller than the restaurant estimate. Direction is a reasonable inference. Size is not.

What we do not know: home services, dental, and fitness

This piece has a real gap at its core. In the research assembled for it, no equivalent large-sample causal study of ratings and revenue was located for home services, dental practices, or fitness. Extending the 5 to 9 percent figure to those verticals is extrapolation, and it should be labeled as such wherever it appears.

It matters to be specific about why the missing evidence cannot simply be replaced with the numbers already circulating in vertical marketing content.

Why the rating point may be worth more now than when it was measured

Luca's natural experiment predates the current discovery environment, and the intervening change runs in one direction: the reputation signal now feeds more than one gate at once. The same aggregate that a human buyer reads is also read by the systems that decide who the buyer sees at all.

The local map pack is the first of these. Local searchers click the local three-pack results far more often than the organic links beneath it, and businesses that appear in the pack draw materially more calls, direction requests, and clicks than comparable businesses that do not. Review count, rating, and recency are among the inputs that influence who occupies those positions, so a rating point does not only persuade the buyer who reaches the profile; it helps decide whether the buyer reaches it.

The generative answer layer is the second gate, and an emerging one. Consumer use of AI tools to find a local business has risen sharply in reported surveys, and AI local recommendations draw heavily on business-profile and review data as their source. A rating point therefore now compounds across surfaces that did not exist when it was first measured. This is a reason to expect the present-day value of reputation to sit at or above the original estimate, and simultaneously a reason not to attach a firmer number to that expectation than the evidence supports.

The signal moves buyers, but trust in the signal is itself moving

A finding that a rating point is valuable is not a license to treat reviews as a lever that pays without limit. The reliability buyers grant the signal is itself shifting, and not only upward.

General trust in online reviews has declined over several years, from a peak near 84 percent to roughly half of consumers in recent survey data. Trust in AI-chatbot output specifically sits lower still: a 2026 survey found only about 29 percent of US chatbot users trust the information a lot or some, even as chatbot use for information search has grown. The practical reading is that reputation work compounds where the signal is credible and consistent, and decays where it looks manufactured. It also underlines why the manufactured shortcut is off the table on evidence grounds as much as legal ones: a reputation the buyer no longer believes is not an asset, whatever its star average.

How to read a "reviews lift revenue by X percent" claim

Because the demand for a clean number is strong and the supply of rigorous ones is thin, the market is full of confident percentages. The useful skill for an owner is not memorizing a figure but grading the claim behind it. Three questions separate an established finding from a borrowed one.

• Is it causal or correlational? A number from a natural experiment or randomized design (Luca's rounding thresholds) measures the effect of the rating. A number from "businesses with more reviews earn more" only shows that quality and reviews travel together.
• Is it the right vertical? The 5 to 9 percent estimate is a restaurant estimate. A source that reports it as the value of a rating point for a dental practice or a gym has transplanted a number across a boundary the original study never tested.
• Is the primary source locatable? An established statistic names an author, a venue, and a year you can check. A figure attributed only to an unnamed "study" or a trade association with no locatable document is a fabrication risk, and several of the most repeated vertical statistics fail this test.

What this means for a high-consideration local business

The evidence supports a precise and limited conclusion. For an independent, owner-operated local business, the kind with no national brand prior doing the persuading, a better reputation signal plausibly carries real revenue weight, and that weight may be larger today than when it was measured because the same signal now governs visibility across the map pack and AI answers as well as the buyer's own reading. The direction is well founded. The exact size, in any vertical other than restaurants, is not yet known and should not be quoted as though it were.

What follows from that is not a promise of a percentage. It is a case for building the reputation signal on real evidence and measuring the movement against a real baseline, so that the value in your own vertical becomes something you observe rather than something you borrow from a restaurant study. That is the difference between a marketing claim and a measured outcome.