A party of eight came in on the Friday your walk in failed. The food was late, one dish went out wrong, and the manager comped a round of drinks. On the Monday the review appeared: three stars, fairly written, factually accurate. Your average moved from 4.6 to 4.5. Nothing else about the business changed. Six weeks later bookings for Wednesday and Thursday were noticeably thinner and nobody could say why.
Restaurant demand does not respond to reviews in a straight line, which is exactly why operators underrate them. The effect arrives late, arrives on your quiet days first, and never announces itself.
The rating is a filter before it is an opinion
Very few people read your reviews. A great many people use your score to decide whether to read anything at all. On Google Maps, on TripAdvisor, inside booking platforms, the first interaction is a filter: show me places above four stars, near me, open now. Falling below a round number does not reduce the number of people who considered you. It reduces the number of people who ever saw you.
That is why the gap between 4.4 and 4.5 behaves so differently from the gap between 3.8 and 3.9. One crosses a rounding boundary in a filter. The other does not.
It also explains the day of week pattern. A booked out Saturday is largely repeat trade and people who decided where they were going days ago. A Wednesday is discovery. Discovery runs through the filter, so the filter hits midweek first.
Arithmetic is why a bad fortnight never quite leaves
Averages are unforgiving. If you have two hundred reviews at 4.6 and you take fifteen one star reviews in a fortnight because a kitchen was short staffed, your average falls quickly and then recovers at the speed of new reviews only. To get back to where you were you need a lot of good ones, and good experiences generate reviews at a much lower rate than bad ones do. That asymmetry is the whole game.
Two practical consequences follow. First, volume is protection. A restaurant with eight hundred reviews absorbs a bad week; a restaurant with forty does not. Second, if you know a period is going to be rough, because your head chef is off or the room is being worked on, the cheapest thing you can do is take fewer bookings rather than accept them and disappoint.
Replying well is a public skill and a private discipline
Your reply is not written for the person who complained. They have gone. It is written for the next reader, who is deciding between you and two other places and who reads your response as a sample of how you behave when things go wrong.
The pattern that works is dull and consistent. Acknowledge the specific thing. Say what happened, briefly and without blaming a member of staff by name. Say what has changed. Offer to continue the conversation off the platform. Stop.
The pattern that fails is equally consistent. Arguing the facts. Implying the guest is lying. Explaining that everyone else was happy that night. Mentioning that they had four bottles of wine. That last one is not only bad manners, it risks a data protection problem, because confirming what someone ate, drank or spent is disclosing their personal data in public. The Information Commissioner's Office takes a dim view and so will every reader.
Resist the urge to threaten legal action. Under the Defamation Act 2013 a statement has to have caused or be likely to cause serious harm to reputation, honest opinion is a defence, and the publicity of trying to silence a diner reliably costs more than the review did.
Asking for reviews is legal, buying them is not
You are allowed to ask. You are allowed to ask everyone, on a card with the bill, on a receipt, in a follow up email if you have the right permission. You are not allowed to ask only the people who looked happy, and you are certainly not allowed to pay for reviews.
The Digital Markets, Competition and Consumers Act 2024 made commissioning, incentivising without disclosure, or publishing fake reviews a banned practice, and it gave the Competition and Markets Authority the power to impose penalties of up to 10% of worldwide turnover directly, without first going to court. The regime came into force in April 2025. Agencies still sell review packages. They are selling you a fine and a story in the local paper.
Incentives are the grey area worth avoiding entirely. Offering a free coffee to anyone who leaves a review, positive or not, must be disclosed by the reviewer, which they will not do. Offering it only for five star reviews is straightforwardly prohibited.
- Ask every table the same way, at the same moment, with the same words
- Make it one tap: a short link or a QR code that lands on the review box
- Never filter who gets asked based on how the meal went
- Reply to the bad ones within a couple of days and the good ones in batches
- Keep a note of what each complaint was actually about, and read it monthly
That last point is where the money is. A review is free research from someone who paid to be there. Fifteen reviews mentioning the temperature by the door is a heating problem worth four hundred pounds to fix, and you would never have found it in your P and L.
The score is a lagging indicator of your worst service
The useful reframing is this. Your rating is not a measure of how good you are. It is a measure of your variance. Nobody writes about consistent competence. They write about the night it went wrong and the night it was unexpectedly excellent, and the first happens more often than the second.
So the lever is not the food at its best. It is the floor under your bad services: enough kitchen cover on the day the delivery is late, a manager empowered to fix things at the table before anyone gets home, and an honest cap on how many covers you take when you are short.
Where this shows up in the game
Restaurants are part of Business Baron's full game, and cafés and pubs are free to run. Reviews accumulate from the service you actually give, and a bad week on staffing shows up in your rating weeks later. How you answer a complaint matters as much as whether you caused it, because care is the slowest part of brand trust to rebuild.