How to find the best-rated businesses of a kind in an area

To find the best-rated businesses of one kind, list every place of that kind in the area, ignore the ones rated by too few people to judge, and rank the rest by rating with the number of reviews as a tie-breaker or a weight. A short top ten built this way is far more trustworthy than sorting by stars alone, where a new place with two perfect reviews jumps ahead of a favourite that hundreds of people have rated just below five.

This is for people who need a shortlist of the best, not a picture of the whole market: a supplier choosing which cafes to approach first, a hotel concierge recommending dentists, a blogger writing a best-of list, a buyer looking for well-regarded independent garages. It explains how ratings and review counts work together, how to set a fair threshold, a made-up worked example, the usual traps, and how to have an AI employee keep the shortlist current.

Why stars alone mislead

A rating is an average. Its reliability grows with the number of people behind it. Three reviews can come from friends on opening day; three hundred are much harder to move. So the question is never only "how high is the rating" but "how high, and how sure".

In many trades high ratings are also crowded at the top, which makes the tie-breaker matter. In xMap's places database for November 2026, of the 366 rated dentists among the 460 nearest the centre of Toronto, 313 are rated 4.5 or more. With that many near the top, the number of reviews decides who is truly best known. In other trades the top is thin: of the 423 rated supermarkets among the 437 nearest the centre of London, 38 are rated 4.5 or more.

313 of 366 ratedToronto dentists rated 4.5 or more

38 of 423 ratedLondon supermarkets rated 4.5 or more

138 of 310 ratedSydney cafes rated 4.5 or more

Three fair ways to rank

• A minimum number of reviews, then by rating. Leave out places with fewer than, say, 20 reviews, then sort by rating. Simple to explain to anyone.
• Rating first, few-review places last. Sort by rating, but put places rated by very few people after all the others. Nobody is excluded, yet nobody climbs on a handful of votes.
• A weighted score. Blend the rating with how many people gave it, so a 4.8 from hundreds can beat a 5.0 from a few. Fairer, but harder to explain in one line.

For a best-of list read by the public, the first or second is usually right, because readers can check it. For an internal shortlist, the weighted score finds better candidates.

Setting the threshold

The right minimum depends on the trade. A busy restaurant gathers reviews fast; a specialist clinic or a wholesaler slowly. A practical way to choose:

1. List all places of the kind in your area with their ratings and review counts.
2. Look at the review count of a place you know is well established. Set the minimum well below it, so similar places pass.
3. Check that the top ten after the cut includes at least some names you would expect. If it holds only unknowns, the minimum is too low; if it holds only chains, it may be too high for independents.
4. Write the rule down, so next month's list is built the same way.

A worked example (made up)

Say a speciality coffee roaster in Cape Town wants the 15 best-rated independent cafes within 5 km of its roastery to offer a tasting. Its first list has 140 cafes. Sorting by stars alone puts four places with one or two reviews at the top. Setting a minimum of 30 reviews leaves 76; sorting those by rating gives a top 15 where every place has between 60 and 900 reviews and a rating of 4.6 or above. The roaster removes two that already buy from it. The figures are invented, but the pattern is common: the cut on reviews changes most of the top of the list.

Mistakes to avoid

• Comparing ratings across trades. Customers rate a beauty salon differently from a supermarket. Rank within one kind of business at a time.
• Mixing areas of very different size. A top ten of a whole city will be dominated by the busiest districts. Rank by district if you need coverage.
• Forgetting closed places. A well-rated place that has closed still has its rating. Leave out places marked closed.
• Treating a rating as fixed. Ratings move. Rebuild the list on a schedule rather than printing it once.
• Copying the list into marketing without checking. Look at each place before calling it "the best" in public.

How to do it with an AI employee

In xMap Flows a shortlist can be rebuilt every month by an AI employee:

1. Schedule: the first Monday of each month.
2. Area of interest: the city or districts, or set a point and a distance in the next block.
3. Nearby places: the one kind you want, with Keep set to Best rated first. It looks at the 500 nearest and keeps the best rated, with each rating weighed by how many reviews it has. Leave Only places that are open switched on, and set a Minimum rating if you want one.
4. Filter: reviews at least your minimum, such as 30.
5. Sort and limit: by Rating, keep the top 15. Places rated by fewer than 10 people come after the others.
6. xMap LLM, optional: one sentence per place on what customers seem to value, from its type, rating and reviews.
7. Email with Show on a map.

What lands each month is a ranked list with name, address, rating, number of reviews, phone and website, plus a map. Add Only new if you want only places that have entered the top list since last month.