How Star Ratings Work (and Why a 4.6 Can Beat a 5.0)
Most star ratings are calculated as a simple average: add up all the stars and divide by the number of reviews. That works well when a product has hundreds of ratings, but it is easy to distort when there are only a few, which is why a single 5-star review can look better than hundreds of 4.6 reviews. On RankTheBrands, the "top rated" sort uses a weighted average that corrects for this, so items with a long track record are not pushed aside by items with one lucky review.
How are star ratings calculated?
The basic method is the arithmetic mean. If a product has four reviews of 5, 4, 4 and 3 stars, the average is (5 + 4 + 4 + 3) / 4 = 4.0.
This simple average has two weaknesses:
- Small samples swing wildly. One review of 5 stars gives a perfect 5.0. One review of 1 star gives a 1.0. Neither tells you much.
- It ignores how many people agree. A 4.8 from five people and a 4.8 from five thousand people look identical, even though the second is far more reliable.
Many sites still sort by this raw average, which is how unknown products with one or two ratings end up at the top of a "highest rated" list.
Is a 5 star rating reliable?
A perfect 5.0 is only reliable when it comes from a large number of reviews. In practice, almost every popular product collects a few lower ratings over time: a damaged delivery, a size that did not fit, a unit that failed early. That is normal.
So when you see a perfect score, check:
- How many reviews it is based on
- Whether the reviews are recent or all from the same week
- Whether the reviews contain real details
If the 5.0 rests on one to three reviews, treat it as "not enough information yet" rather than "the best". Our guide on how to spot fake reviews covers more warning signs.
What is a Bayesian average rating?
A Bayesian average rating adds a small number of imaginary "starting" reviews to every item before calculating the average. These starting reviews pull items with few ratings towards a neutral middle value. As real reviews come in, they quickly outweigh the starting reviews, and the score moves towards the true average.
The idea is simple: until there is enough evidence, assume an item is about average. Once many people have rated it, trust what they say.
The formula used on RankTheBrands
The "top rated" sort on RankTheBrands uses this weighted average:
Weighted rating = (C · m + Σ ratings) / (C + n)
Where:
- C = 5, the number of imaginary starting reviews
- m = 3.5, the star value of each starting review
- Σ ratings is the sum of all real star ratings the item has received
- n is the number of real reviews
In plain words: every item is treated as if it started with 5 reviews of 3.5 stars. The term C · m is therefore 5 × 3.5 = 17.5 stars spread across 5 imaginary reviews.
A worked example
Imagine three products in the same category.
Product A has 1 review of 5 stars.
- Simple average: 5.0
- Weighted: (17.5 + 5) / (5 + 1) = 22.5 / 6 = 3.75
Product B has 10 reviews, all 5 stars.
- Simple average: 5.0
- Weighted: (17.5 + 50) / (5 + 10) = 67.5 / 15 = 4.50
Product C has 100 reviews with an average of 4.6 stars, so the ratings add up to 460.
- Simple average: 4.6
- Weighted: (17.5 + 460) / (5 + 100) = 477.5 / 105 ≈ 4.55
With a simple average, A and B tie at the top with a perfect 5.0 and C comes last. With the weighted average, C comes first, B is a close second, and A drops to third. That matches common sense: a hundred people broadly agreeing on 4.6 is stronger evidence than one enthusiastic review.
Notice also how quickly real reviews take over. With 100 reviews, the imaginary starting reviews shift C's score by only about 0.05 stars. With one review, they dominate.
Why use 3.5 as the starting value?
The starting value should sit a little below the typical rating for well-reviewed products. Online ratings tend to cluster in the upper half of the scale, so 3.5 is a cautious middle point: new items are not punished harshly, but they also cannot jump ahead of established items with a single perfect score. If a new product really is excellent, it will rise as more people rate it.
How to read ratings as a shopper
- Look at the number of reviews next to the stars, not just the stars.
- Prefer a slightly lower score from many people over a perfect score from a few.
- Read the middle ratings (three and four stars) for balanced pros and cons.
- Compare products within the same category, such as personal care or home appliances, because expectations differ between categories.
- Check whether recent reviews differ from older ones, which can signal a change in quality.
Frequently asked questions
How are star ratings calculated on most websites?
Most websites use a simple average: the total number of stars divided by the number of reviews. Some, including RankTheBrands for its "top rated" sort, use a weighted average that accounts for how many reviews an item has.
What is a Bayesian average rating in simple terms?
It is an average that adds a few imaginary neutral reviews to every item. This keeps items with very few ratings from appearing at the extremes until more real reviews arrive.
Is a 5 star rating reliable?
Only when it comes from a large number of reviews. A perfect score from one or two people tells you very little.
Why does a product with a 4.6 rank above one with a 5.0?
Because the 4.6 is based on many more reviews. On RankTheBrands, the weighted average gives more confidence to scores backed by a larger number of ratings.
Does the weighted score change the stars shown on a product?
The weighted average is used to order items in the "top rated" sort. The number of reviews is shown alongside the rating so you can judge it yourself.
Add your rating
Every review makes the scores on RankTheBrands more accurate. Browse all categories, read what other users think, and add your own rating for the products you know well.