Restaurant review analysis: what your reviews tell you

Your star average is the slowest number on your listing: it is built from every review you have ever had, so a new problem can run for months before it moves. Reading your reviews by topic, month by month and next to the places nearby shows which complaint is growing now. Here is how to do it with a spreadsheet, worked on a made-up example, plus what research says reviews are worth and how to ask for more within the rules.

What the research says reviews do to sales

Two studies get quoted most. Neither is a promise about your restaurant.

• Revenue. Michael Luca's Harvard Business School working paper 12-016 found that a one-star rise in a restaurant's Yelp rating went with a 5 to 9% rise in revenue. The effect came from independent restaurants; chains showed none. The sales data came from the Washington State Department of Revenue, and the paper first came out in 2011, so the data is well over a decade old.
• Full tables. Michael Anderson and Jeremy Magruder, in the Economic Journal in 2012, looked at San Francisco restaurants' 7 p.m. reservations. In their own summary, an extra half-star on Yelp made a restaurant sell out 19 percentage points (49%) more often at peak hours, with larger effects where guests had less other information. For restaurants with more than 500 reviews they found no jump.

Both are careful studies, but each covers one city or state, years ago, on Yelp, not Google. The fair reading: what guests see about you can move how many of them come, most of all when they know little else about you.

There is no reliable list of top complaints

Search for what restaurant guests complain about most and you will find neat percentages. We looked for where they come from and found no traceable source: no government survey, no large peer-reviewed count of US restaurant reviews. The academic studies read narrow samples, such as fine dining or one city, and the rest comes from companies that sell review software. What guests complain about differs from one restaurant to the next, which is the reason to count your own.

Read by topic, not by stars

A 3-star review can praise the food and complain about the wait in the same breath. Sorting by stars hides that; sorting by topic shows it. Eight topics cover most of what guests write about: food; service and staff; wait times; price and value; cleanliness; atmosphere; order accuracy; takeout and delivery.

Open a spreadsheet with one row per review: the date, the stars, and two columns per topic, praise and complaint. Read each review once and mark every topic it touches. A hundred reviews take an evening.

The food is the café's strength. Wait times are the surprise: the most complaints of any topic, and almost no praise to balance them.

Month by month: is it growing?

A total doesn't say whether a problem is fading or growing. Split the same reviews by month.

Wait complaints went from 1 review in 16 in June to more than 1 in 5 in October, while the month's average fell from 4.5 to 4.1. The café's overall rating, built from years of reviews, would hardly move. A second check: compare your newest 25 reviews with the 75 before them. A topic much worse in the newest 25 is recent.

Then look for the cause in your own records: a schedule change, someone who left, busier lunches. The sales per labor hour guide shows whether a shift is short-staffed for its sales, and our post on slow shifts covers moving hours to where the rush is. If value complaints grow after a menu change, see how much to raise menu prices. Wrong orders often come down to one habit: reading the order back.

Compare with the places nearby

Some complaints belong to the whole block: when prices rise everywhere, every place gets more value complaints. Read three similar places near you the same way and compare the share of reviews, not the count.

Price complaints are common on this block, and the café has fewer than its neighbors. The wait is the café's own: more than twice the neighbors' average. Read what they are praised for, too.

How many reviews make a pattern?

There is no fixed number, and anyone who gives you one is guessing. Some rules of thumb:

• One angry review is one guest's night. Answer it, but don't rebuild a menu around it.
• The same complaint from different guests in different weeks is worth checking on the floor.
• A complaint whose share rises month after month matters more than a large total that is fading.
• With few reviews a month, one review moves the share a lot. Treat small numbers as a hint and check your records first.

Asking for more reviews within the rules

More reviews make your monthly counts steadier. The rules differ by who sets them.

The FTC. The Federal Trade Commission's rule on fake reviews bans fake reviews, including AI-generated fake ones, undisclosed reviews by owners and managers, and scaring guests into taking reviews down. Under section 465.4, a business may not offer anything for a review that must be positive or negative. The FTC's questions and answers say the rule took effect on October 21, 2024, and that asking only happy guests, while not banned by the rule itself, could still break the wider FTC Act.

Google is stricter. Its Maps content policy does not let a business:

• offer any reward, such as a discount or a free dessert, for posting, changing or removing a review;
• discourage negative reviews, or ask only the guests it expects to be happy;
• pressure guests to review while still in the restaurant, or ask for specific content;
• ask staff to collect a certain number of reviews, or reviews that mention a staff member.

Reviews from current or former employees are removed as a conflict of interest. What Google does suggest, in its tips to get more reviews, is asking customers to visit a Google link or scan a QR code. Ask every guest the same way: a card with the check, a line on the receipt, a sign by the door.

Yelp asks the opposite. Its page “Don't Ask for Reviews” says its software may not recommend reviews that seem prompted. So point your sign to Google, not Yelp.

If you'd rather not count by hand

Our review reader reads your newest 100 Google reviews and counts praise and complaints across the same eight topics, sets each topic's share beside up to three places like yours nearby, and shows month by month how many reviews complain and their average stars, with your newest 25 reviews' stars against the rest when there are 50 or more. It lists staff guests praise by name (praise only) and drafts replies to unanswered 1 to 3 star reviews for you to edit and post.

It reads Google only, never posts or asks for reviews, and you paste nothing. A “since your last reading” comparison comes only with Review Watch, which also checks your 20 newest reviews each week and emails a reply draft for each new, unanswered 1 or 2 star review that has words (up to five a week). See a sample result first.

This week

1. 01Mark your newest 50 to 100 Google reviews in a spreadsheet: eight topics, praise and complaint for each.
2. 02Split them by month and find the one topic whose complaints are rising.
3. 03Read three similar places nearby the same way, and compare shares, not counts.
4. 04Check the rising complaint against your own schedule, tickets and sales for those weeks before you change anything.
5. 05Put a Google review link or QR code where every guest sees it, and drop any reward, staff target or request to name a server.