Section / 6 pages
The math of star ratings: small samples, intervals and averages
Six short pieces that show their working. Every table can be recomputed by hand, every chart is drawn from the numbers on the page, and each example says which figures are invented for illustration.
A rating is a statistic, and most confusion about ratings is confusion about statistics: reading a small sample as if it were a large one, treating a rounded display as exact, forgetting how hard an average is to move once it has history behind it.
Start with the worked comparison of 4.8 from 12 reviews against 4.6 from 900. The calculator lets you run the same method on any profile. None of this arithmetic deals with fake or selectively gathered reviews; for that side, see the explainers.
- Why 4.8 from 12 reviews is weaker evidence than 4.6 from 900WORKED EXAMPLEA worked example: a 4.8 from 12 reviews is consistent with a true average from 4.4 to 5.0. A 4.6 from 900 is pinned between 4.53 and 4.65.
- A simple confidence-interval explainer for star ratingsMETHODA confidence interval turns "4.6 stars" into "probably between 4.4 and 4.8". How to compute one for an average and for a percent-positive score.
- Star rating confidence calculator, with a worked tableCALCULATOREnter how many 5, 4, 3, 2 and 1 star reviews a business has. Get the mean, a 95% range and a cushioned average. Static worked table included.
- Bayesian averages: what seven phantom reviews do to a ratingWORKED EXAMPLETrustpilot adds seven neutral 3.5-star reviews to every score. A worked look at how that prior pulls small profiles to the middle and fades with volume.
- Rounding and the half-star cliff: how 0.01 can change the displayWORKED EXAMPLEWhen ratings round to the nearest half star, 4.24 displays as 4 stars and 4.25 as 4.5. Worked examples of what it takes to cross a threshold.
- How much does one review move an average rating?WORKED EXAMPLEOne 1-star review drops a 4.8 average by 0.35 at 10 reviews and 0.04 at 100. Undoing it takes 19 five-star reviews at any size. The arithmetic.