RATINGS/DECODED

Explainers / READING

What a star rating can and cannot tell you about a local business

A star rating is a summary of the people who chose to leave a review on one platform, processed by that platform's rules. It can tell you roughly how past reviewers felt. It cannot, by itself, tell you about workmanship, licensing, price, or how the next job will go.

Start with what the number is. On Google, the review score is "the average of all ratings published on Google for that place or business".11 Google Business Profile Help, Understand review scores for local places and businesses On Trustpilot it is a weighted figure that favors recent reviews and quietly includes seven neutral ones.22 Trustpilot Help Center, TrustScore and star rating explained On Amazon it comes from "machine-learned models instead of a simple average".33 Amazon Customer Service, Understanding Customer Reviews and Ratings On the BBB, the letter grade is not built from customer reviews at all.44 Better Business Bureau, Overview of BBB Ratings Four well-known scores, four different objects. Before asking whether 4.7 is good, it helps to ask what kind of thing 4.7 is on the page in front of you.

4.8 from 12 reviews 4.6 from 900 reviews

What the number does tell you

How the people who reviewed felt, on average. That sounds obvious, and it is the full extent of the direct claim. The reviewers are a subset of customers, and they picked themselves. Research that pooled more than 280 million reviews from 25 platforms found that most review distributions are polarized, with many reviews at the positive end, few in the middle and some at the negative end, and traced much of this to who decides to write.55 Schoenmueller, Netzer and Stahl, Journal of Marketing Research 57(5), 2020, The Polarity of Online Reviews: Prevalence, Drivers and Implications The average of a self-selected, lopsided group is still information. It is information about that group.

Whether something has gone badly wrong, often. A rating well below its local competitors, built from hundreds of reviews, is hard to produce by accident. Large samples of unhappy reviewers usually mean a real pattern, and reading a dozen of the low reviews will tell you which one: missed appointments, billing disputes, rude staff.

Roughly where a business sits relative to others on the same platform. Ratings work better as a ranking tool inside one site than as an absolute grade. A 4.3 on Yelp and a 4.3 on Google were produced by different filters and different crowds.

What the number cannot tell you

It cannot tell you about technical quality

For products, this has been tested. A study of 1,272 products across 120 categories found that average user ratings "lack convergence with Consumer Reports scores", were often based on too few ratings to be informative, and did not predict resale prices.66 de Langhe, Fernbach and Lichtenstein, Journal of Consumer Research 42(6), 2016, Navigating by the Stars: Investigating the Actual and Perceived Validity of Online User Ratings The same study found ratings ran higher for expensive products and premium brands even after controlling for the test-lab scores.

No equivalent lab exists for roofers or dentists, which makes the problem harder for services, not easier. A homeowner can judge whether the crew showed up on time and whether the invoice matched the quote. She usually cannot judge whether the flashing was installed correctly. Reviews of local services are mostly reviews of the customer experience: communication, punctuality, friendliness, price shock. Those things matter. They are a different question from whether the work will hold up in six years.

It cannot tell you whether the business is licensed or insured

Nothing in a star average checks a license. Some platforms run their own screening, and they say so on their own pages. Angi describes a background check on the owner or principal for businesses in its paid "Angi Approved" tier, and says the business "must attest" to holding applicable licenses.77 Angi, Frequently Asked Questions Thumbtack shows a "Licensed pro" badge when it has "used a public database to verify the license number the pro provided".88 Thumbtack, How we're committed to your safety These are separate signals from the stars, with their own limits, and the state licensing board's own lookup remains the authority.

It cannot tell you how sure to be

A rating printed as 4.8 looks equally solid whether it rests on 12 reviews or 12,000. It is not. With 12 reviews, one unhappy customer moves the average by about three tenths of a star. We work a full example in the math section, where a 4.8 from 12 reviews turns out to be compatible with anything from about 4.4 to a perfect 5, while a 4.6 from 900 is pinned between roughly 4.53 and 4.65. See why 4.8 from 12 reviews is weaker than 4.6 from 900.

The research on shoppers is blunt about how people handle this. Consumers "fail to moderate their reliance on the average user rating as a function of sample size sufficiency".66 The authors call the resulting confidence an "illusion of validity". de Langhe, Fernbach and Lichtenstein, Journal of Consumer Research 42(6), 2016, Navigating by the Stars: Investigating the Actual and Perceived Validity of Online User Ratings People read the average and skip the count.

It cannot tell you which reviews were left out

Every platform removes or sidelines some reviews, and the rules differ. Yelp's software decided that 17% of the 22 million reviews contributed in 2025 would not be recommended, and its staff removed another 11%; neither group counts toward the star rating.99 Yelp Trust and Safety, 2025 Trust and Safety Report Google says it blocked or removed more than 170 million policy-violating reviews in 2023.1010 Google, The Keyword, How machine learning keeps contributed content helpful Filtering is necessary, because fraud is real, but it means the displayed average is the average of what survived. See why reviews disappear.

It cannot tell you how the reviews were collected

A business that asks every customer for a review will have a different profile from one that never asks, and both will differ from one that only asks the customers who seemed pleased. The last practice is against Google's rules, which bar merchants from choosing to "selectively solicit positive reviews from customers".1111 Google Maps Help, Prohibited and restricted content, Maps user-contributed content policy It still happens, and from the outside you mostly cannot see it. Trustpilot is unusual in showing readers whether a business invites reviews.22 Trustpilot Help Center, TrustScore and star rating explained

A workable way to use the number

Treat the star rating as the first filter and never as the decision.

  1. Read the count before the average. Under about 30 reviews, treat the first decimal place as noise. Our piece on how many reviews are enough gives the reasoning.
  2. Sort by newest. A business can change owners or crews, and several platforms weight recent reviews for this reason. The last ten reviews describe the business you would be hiring.
  3. Read the low reviews for facts, the high ones for specifics. A one-star review that names a date, a job and a dollar figure is evidence. A five-star review that says only "great service" is weak evidence of anything.
  4. Read how the business answers. A reply that addresses the complaint tells you how disputes get handled. This is one of the few places a rating page shows the business speaking for itself.
  5. Check a second platform. Agreement between two sites with different filters is worth more than a high score on one.
  6. Verify the things stars do not measure. License status with the state board, insurance by certificate, references for large jobs.

Why bother with ratings at all

Because they move behavior and, on balance, they carry signal. In a 2016 Pew Research Center report, 82% of U.S. adults said they at least sometimes read online ratings or reviews before buying something for the first time, yet only about half of review readers said reviews generally give an accurate picture.1212 Pew Research Center, Online Shopping and E-Commerce, section 2: Online reviews (December 2016) Readers already sense the gap between popularity and proof. The aim of this site is to make that instinct specific: to say which parts of the number are arithmetic, which parts are policy, and which parts the platform does not disclose.

Questions readers ask

Is a higher star rating always better?

No. Northwestern University's Spiegel Research Center reports that, for products, the likelihood of purchase tends to top out somewhere from 4.0 to 4.7 stars and then declines toward 5.0.1313 Medill Spiegel Research Center, Northwestern University, How Online Reviews Influence Sales Shoppers appear to read a flawless score as less credible. A 4.6 with many detailed reviews is often a stronger sign than a 5.0 with a handful.

Does a low rating mean a business does bad work?

It means its reviewers were unhappy, which is related but distinct. Low ratings often track service problems such as scheduling and billing more than technical skill. Read the text of the low reviews to see which it is, and check whether the complaints are recent.

Are ratings on one platform comparable with ratings on another?

Only loosely. Each platform has its own formula, its own filter and its own set of reviewers. A study comparing hotel reviews on a site where anyone could post with a site limited to verified customers found systematic differences in the mix of positive and negative reviews for the same hotels.1414 Mayzlin, Dover and Chevalier, American Economic Review 104(8), 2014, Promotional Reviews: An Empirical Investigation of Online Review Manipulation

Method We read each platform's published description of its headline number, then set those descriptions beside the peer-reviewed research on what ratings measure. Nothing here comes from testing platforms with fake accounts. See how we source and check.

Sources

Every source below was opened and read for this page. Links go to the publisher's own copy.

  1. Google Business Profile Help. Understand review scores for local places and businesses
  2. Trustpilot Help Center. TrustScore and star rating explained
  3. Amazon Customer Service. Understanding Customer Reviews and Ratings
  4. Better Business Bureau. Overview of BBB Ratings
  5. Schoenmueller, Netzer and Stahl, Journal of Marketing Research 57(5), 2020. The Polarity of Online Reviews: Prevalence, Drivers and Implications
  6. de Langhe, Fernbach and Lichtenstein, Journal of Consumer Research 42(6), 2016. Navigating by the Stars: Investigating the Actual and Perceived Validity of Online User Ratings
  7. Angi. Frequently Asked Questions
  8. Thumbtack. How we're committed to your safety
  9. Yelp Trust and Safety. 2025 Trust and Safety Report
  10. Google, The Keyword. How machine learning keeps contributed content helpful
  11. Google Maps Help. Prohibited and restricted content, Maps user-contributed content policy
  12. Pew Research Center. Online Shopping and E-Commerce, section 2: Online reviews (December 2016)
  13. Medill Spiegel Research Center, Northwestern University. How Online Reviews Influence Sales
  14. Mayzlin, Dover and Chevalier, American Economic Review 104(8), 2014. Promotional Reviews: An Empirical Investigation of Online Review Manipulation