Data in the wild · 4 min read · Sep 26, 2026

The five-star restaurant might not be your restaurant

You’re in a neighbourhood you don’t know well and want a good dinner. Two places are nearby.

Which would you choose?

The first has the higher average. The second has far more evidence. Neither number tells you whether you would enjoy the evening.

That is the problem with treating a rating as a verdict. It is a summary of other people’s experiences, collected under conditions you cannot fully see.

Start with how much the number can move

The 4.9-star restaurant may be wonderful. It may also be new, with its first reviews coming from people especially excited to visit it.

With only 12 reviews, a few different experiences could change the average substantially. An 800-review average is less easily moved by one new opinion, although it may include years of visits under different menus, staff and ownership.

So review count changes how I read a rating. It does not settle the choice.

I would look at the dates. Are most of those 800 reviews from several years ago, or do recent diners still describe the place in the same way? Did the 4.9-star restaurant receive its reviews across several months, or almost all at once?

A large history helps only if it still describes the restaurant that exists today.

Ask who felt moved to review

Customers are not randomly selected and asked for an equally detailed report after dinner. People choose whether to post, what to mention and which platform to use.

Researchers have examined how that selection—and the platform itself—can shape online ratings. A study on review bias discusses influences beyond a simple average, while research on platform selection and measurement shows why reviews collected in different places should not automatically be treated as interchangeable.

That does not make reviews useless. It means a rating describes the people who reviewed, not a controlled survey of everyone who ate there.

The written reviews help you understand the sample. Perhaps one-star reviews mostly concern delivery delays, while you plan to dine in. Perhaps five-star reviews praise the room and service, while saying little about the food.

The same average can hide very different evenings.

Look for the experience you actually want

“Best restaurant” is an incomplete question.

Best for a quick meal before a show? A long conversation over wine? Trying a dish you cannot easily find at home? Sitting outside and watching a neighbourhood go by?

Someone who wants a lively, crowded dining room may award five stars to the very place another person finds too loud. Neither is wrong. They are measuring the experience against different hopes.

This is where review text becomes more useful than another decimal point. Look for repeated, concrete observations that matter to your plan: pace of service, noise, whether people linger, what the kitchen does particularly well, and how the place feels at the time you intend to go.

Give extra attention to reviews that describe trade-offs. “The food was excellent, but the meal felt rushed” tells you more about fit than “Amazing, five stars.”

Keep the neighbourhood in the decision

A restaurant is also part of an evening.

A slightly lower-rated place on a street you want to explore may give you a better night than a higher-rated one that requires crossing town for a single reservation. You might prefer to walk, stop somewhere that catches your eye and leave room for a second glass of wine or dessert.

That cannot be captured in the restaurant’s average rating because the rest of the evening is yours.

This is a small product lesson, too: a metric can be accurate about the thing it measures and still fail to answer the decision someone is making.

A star rating is a useful starting point. The interesting part is asking what produced it—and whether those experiences resemble the one you want.