Is 4.8 Stars Really Better Than 4.3? How to Read a Star Rating Like a Statistician
A perfect score from three people tells you less than a 4.3 from nine hundred. Here is the simple maths review sites use behind the scenes, and a 30-second routine for reading any rating.
Updated October 30, 2026
Two serums sit side by side. One shows 5.0 stars from 3 reviews. The other shows 4.3 stars from 900 reviews. Most of us feel the pull of the 5.0, and that instinct is usually wrong. The average is only half of a rating. The other half is how many people it is built on, how their scores are spread, and how recent they are. None of this needs a calculator. It needs four questions.
1. How many reviews is the average built on?
Small samples swing wildly. With 3 reviews at 5 stars, a single 1-star review drops the average to 4.0. With 300 reviews averaging 4.5, that same 1-star review moves it by about 0.01. A small count does not mean a product is bad. It means the number is not settled yet, so it deserves less weight than it appears to have.
The opposite mistake exists too. In a 2017 study published in Psychological Science, people leaned toward the product with more reviews even when its low average made it the likelier disappointment. Neither number works alone: read them together.
2. The trick review sites use: the Bayesian average
Ranking sites solve the small-sample problem by imagining that every product starts with a handful of average ratings already in place. A product with few real reviews stays close to that typical score; a product with hundreds of reviews has earned its own number. IMDb has long published a weighted formula of this kind for its Top 250 films, and the same idea underpins many "best of" rankings.
You can do a rough version in your head. Suppose a typical product in the category scores 3.8, and you pretend each product already has 10 ratings at that level:
| Product | Shown rating | Reviews | Adjusted rating |
|---|---|---|---|
| Serum A | 5.0 | 3 | (3 × 5.0 + 10 × 3.8) ÷ 13 ≈ 4.08 |
| Serum B | 4.3 | 900 | (900 × 4.3 + 10 × 3.8) ÷ 910 ≈ 4.29 |
Once both numbers are on the same footing, Serum B comes out ahead. Serum A might still turn out to be excellent, but right now the evidence for it is three people.
3. What shape are the stars?
Online ratings rarely form a neat bell curve. Researchers studying review sites have described a J-shaped pattern: lots of 5-star reviews, a smaller pile of 1-star reviews, and few in between, because people with strong feelings are the ones who bother to write. That is why the star breakdown matters more than the average.
- Mostly 4s and 5s, a thin tail of 1s: a product most people are happy with. Read the 1-star reviews to see whether their reason applies to you.
- Two humps, at 5 and at 1: a polarising product. It works brilliantly for some people and fails for others. Find out which group you are in: skin type, hair type, taste, use case.
- Bunched in the middle, around 3: consistently okay. Fine if the price is right, rarely anyone's favourite.
Two products can both average 3.0 and be completely different buys. One is reliably average; the other is a coin flip.
4. How fresh are the reviews?
Formulas change, factories change, packaging changes. A product that earned its rating three years ago may not be the one on the shelf today. Sort by newest and check whether recent reviews agree with the old ones. A sudden burst of glowing reviews in a few days is a different problem, covered in our guide on how to spot fake reviews.
Why honest platforms show lower numbers
If you are used to seeing 4.7 everywhere, an honest platform can look harsh. When nobody can buy, seed or remove ratings, scores spread out and averages drop. On a platform like that, a 3.9 with a healthy breakdown can be a genuinely good product, and a 4.8 is rare and meaningful. Compare scores within the same platform, never across platforms.
The 30-second routine
- Check the count before the average. Under a couple of dozen reviews, treat the average as a first impression.
- Pull it toward the middle in your head: few reviews means "probably closer to typical".
- Look at the breakdown. Two humps means read the 1-star reviews carefully.
- Sort by newest to see whether the product today matches its reputation.
- Read three reviews from people like you. Same skin type, same use, same budget.
Ratee shows all of this on one screen: scan a barcode and you get the community score, the number of ratings behind it, the full star breakdown and the reviews other members found most helpful, with no paid placements.
Scan a product and see the full breakdown, not just the average.
Frequently asked questions
- Is a 5-star rating with only a few reviews trustworthy?
- It is a weak signal rather than a strong one. With only a handful of reviews, one more opinion can move the average a lot, so a perfect score from three people says little about how most buyers will feel. Weigh it against products with more reviews and read what those few reviewers actually wrote.
- How many reviews does a product need before its rating is reliable?
- There is no magic number, but averages built on a couple of dozen reviews or fewer still move noticeably with each new rating. Past roughly a hundred reviews, a single new opinion barely changes the score, so the shape of the breakdown and the recency of reviews become more useful than the count.
- What is a Bayesian average rating?
- It is an average that starts every product with a few imaginary ratings at a typical score for its category, then adds the real ones. Products with few reviews stay close to typical; products with many reviews end up near their own average. It stops a 5.0 from three reviews outranking a 4.3 from hundreds.