Average Rating Calculator
Every star rating you have ever trusted is a weighted average: each star level multiplied by how many reviews gave it, divided by the total review count. It is simple, and it breaks in exactly two ways: lopsided distributions that hide behind a respectable mean, and tiny review counts that make a 5.0 out of two reviews look better than a 4.6 out of two thousand.
The Average Rating Calculator below computes the weighted average and then applies the fix the big platforms actually use: the Bayesian average, which blends your rating toward a prior until you have enough reviews to stand on your own. Enter your review counts and see both numbers, plus the distribution chart.
Quick Answer
Key takeaways
- Average rating is a weighted mean: star value × count, summed, divided by total reviews.
- The Bayesian average blends your rating toward a prior so small samples cannot outrank established ones.
- Two products can share a 3.0 average with nothing else in common: check the distribution shape, not just the mean.
- Report the review count next to every average. A number without its sample size is a rumor.
The weighted mean behind every star rating
The formula first, plainly: multiply the number of 5-star reviews by 5, 4-star by 4, and so on, add those five products, and divide by the total number of reviews. Each review counts once, but a 5-star review contributes five times as much to the numerator as a 1-star review. That is what “weighted” means here: the star value is the weight.
With the defaults: (5×120) + (4×80) + (3×30) + (2×15) + (1×5) = 600 + 320 + 90 + 30 + 5 = 1,045, divided by 250 reviews = 4.18 stars. As a share of the 5-point maximum that is 83.6%, handy when you compare against a 10-point scale elsewhere.
Worked example, checked against this exact tool
Enter the pre-filled defaults and click Calculate. The distribution chart above is drawn from your numbers live, and the results read:
| Figure | Value |
|---|---|
| Total reviews | 250 |
| Weighted average | 4.18 stars |
| Bayesian average | 4.07 stars |
| Share of maximum | 83.6% |
With 250 reviews the Bayesian pull is gentle: 4.18 slides to 4.07. Now try 2 five-star reviews and nothing else: the weighted average screams 5.00 while the Bayesian average with the same prior reports 3.15. That gap is the entire point of the next section.
Why small review counts lie: the Bayesian fix
A 5.0 average from two reviews carries almost no information; a 4.3 from a thousand carries a great deal. Ranking by raw average therefore promotes every newcomer with two friendly reviews above every proven product, which is why large platforms blend ratings toward a global mean until enough reviews accumulate. The standard formula:
Think of m as imaginary reviews at the prior average mixed into your real ones. With m = 25 and C = 3.0, two perfect reviews become (2×5 + 25×3) / 27 = 3.15: good, not chart-topping. At 250 reviews the same prior barely dents the 4.18. Raise m if your category is spam-prone; lower it if you want newcomers to climb faster. The calculator lets you tune both C and m to your platform’s reality.
The bimodal warning
The second way averages mislead has nothing to do with sample size. Fifty 5-star reviews plus fifty 1-star reviews average exactly 3.0, and so does one hundred straight 3-star reviews. The first product is polarizing: people love it or return it. The second is uniformly mediocre. Same average, opposite businesses.
Bimodal splits are common where a product works brilliantly for one group and badly for another: think a game that hardcore fans adore and casual players bounce off. The calculator’s verdict badge flags polarized distributions (heavy 5-star and 1-star tails), and the distribution chart makes the shape visible at a glance. Never compare two 3.0s without looking at what built them. The same weighting idea powers survey analysis and portfolio math wherever groups differ in size.
What competing calculators miss
The competent calculators in this space compute the weighted average correctly, and one of them goes further with a Wilson score lower bound, which is genuinely useful statistics. But none of them implements the Bayesian average, the adjustment that actually decides rankings on real platforms, and none draws the distribution. You get a number with no sense of its shape or its trustworthiness.
This tool gives you all three views: the weighted average everyone expects, the Bayesian average with tunable prior for ranking decisions, and the live distribution chart with a polarization warning. The average answers “how good”; the Bayesian answers “how sure”; the chart answers “what kind of good.”
Frequently asked questions
Multiply the count at each star level by that star value, add the five products, and divide by the total number of reviews. For 120 five-star, 80 four-star, 30 three-star, 15 two-star and 5 one-star reviews, that is 1,045 / 250 = 4.18 stars.
It blends an item’s observed average toward a prior (usually the platform-wide average) until enough reviews accumulate to outweigh it: (v×R + m×C) / (v + m). It stops products with two perfect reviews from outranking proven ones, which is why large review platforms use it for rankings.
Set C to your platform or category’s overall average rating, often around 3.5 to 4.0 for commerce sites. Set m to the review count at which you start trusting an item’s own average: 25 is a common default; higher values are more conservative against review spam.
Because the distribution shape matters. Fifty 5-star plus fifty 1-star reviews averages 3.0, and so does one hundred 3-star reviews, but the first is polarizing and the second is mediocre. Always read the distribution alongside the mean.
For display, yes: one decimal (4.2 stars) is the platform convention and avoids implying false precision. Keep full precision internally for ranking math like the Bayesian blend.
It delays them, which is the point: a newcomer must earn its rank with volume. With sensible m values the penalty fades fast; at 10× m reviews the observed average dominates and the prior is a rounding error.
Add This Calculator to Your Website
References
- Evan Miller, “How Not To Sort By Average Rating” – the standard exposition of the Bayesian average formula (v×R + m×C)/(v + m) for ranking items. Evan Miller: How Not To Sort By Average Rating. Accessed October 19, 2026.
- District Data Labs, “Computing a Bayesian Estimate of Star Rating Means” – worked Bayesian estimation on movie ratings showing the prior’s pull at low counts. District Data Labs: Bayesian Estimate of Star Ratings. Accessed October 19, 2026.
- OnlyCalculators, “Average Rating Calculator” – the weighted mean formula and the Wilson score lower bound approach. OnlyCalculators: Average Rating Calculator. Accessed October 19, 2026.
- Dash Calculator, “Star Rating Calculator” – the bimodal distribution problem and why sample size matters for rating trust. Dash Calculator: Star Rating Calculator. Accessed October 19, 2026.
- TrueGeometry Blog, “Rating Algorithms” – Bayesian average pros, cons, and the m parameter’s role. TrueGeometry: Rating Algorithms. Accessed October 19, 2026.