The most quoted number in online reputation comes from a 2011 Harvard Business School working paper by Michael Luca, revised in 2016. It found that a one-star increase in a restaurant's Yelp rating led to a 5 to 9% increase in revenue. That figure gets pasted into sales decks constantly, usually without the second half of the finding, which is the more useful one.
The effect appeared only for independent restaurants. For chain restaurants it was not detectable at all.
Why the study is worth trusting
Most claims about ratings and revenue are correlations dressed up as causes. Good restaurants earn good ratings and also earn good revenue, so finding the two together proves nothing about whether the rating did any work.
Luca got around this by exploiting an accident of how Yelp displays scores. Yelp rounds to the nearest half star, so a restaurant averaging 3.24 displays 3.0 and one averaging 3.26 displays 3.5. Those two restaurants are, for practical purposes, identical. The only difference between them is which side of a rounding threshold they landed on, and therefore what number a customer sees. Comparing revenue across that threshold isolates the effect of the displayed rating itself.
That design is why the number survived peer scrutiny and why it still gets cited fifteen years later.
The chain result is the important one
A customer choosing between two unfamiliar independent restaurants has almost nothing to go on. They have never eaten at either. Nobody they know has an opinion. The rating is close to the only signal available, so it carries the decision.
A customer choosing a chain already knows what they are getting. They have eaten at one before, or seen a hundred of them, or watched the advertising. The brand has already answered the question the rating exists to answer, so the rating stops mattering.
Reviews do not create trust out of nothing. They substitute for reputation you have not had time to build. The less a customer already knows about you, the more the rating decides.
Read that way, the finding tells you exactly who should care most:
- A single location, or a young brand, where most customers are first-timers.
- A new outlet in a city where the brand has no presence yet.
- Any business competing against nearby alternatives a customer considers interchangeable.
- Categories bought infrequently, where nobody has a personal history to fall back on.
And who should care less: an established national brand whose customers already know the answer before they search.
What this does not say
Three limits worth stating plainly, because the number is routinely stretched past all of them.
It is restaurants, in the United States, on Yelp. The mechanism travels well and the specific percentage does not. Anyone quoting "5 to 9%" for a dental clinic in Pune is extrapolating, and should say so.
It is a one-star move, which is enormous. Going from 3.5 to 4.5 is not a quarter of review management. It is years of operational change that reviews merely record. Most businesses working seriously at this move a few tenths, and a few tenths is worth proportionally less.
The rating is a readout, not a lever. You cannot push it directly. It moves when the underlying experience moves, or when the set of customers who bother to leave a review changes. Everything reputation work can legitimately do falls into that second category: making it normal for satisfied customers to say so, and dealing with problems before they become the only thing on record.
What to do with the number
Use it to decide where attention goes, not to forecast revenue. The honest version of the argument is this: if your customers are mostly choosing you without knowing you, your rating is doing more work than any other single thing on your listing, and it is the cheapest thing on that listing to improve.
If your customers already know exactly what they are buying, spend the effort elsewhere. The research is fairly clear that it will not pay you back.