It is the most common shortcut in online shopping: sort by rating, then trust whichever product has the most reviews behind it. We checked that shortcut against 30,470 rated listings. It does not do what most people assume — and the reason is more useful than the answer.
No. More reviews go with less separation between products, not more.
Among the listings we track with more than 10,000 reviews, 98.2% are rated 4.0★ or above. The median rating is 4.30 — and it is 4.30 at every review threshold we tested. Filtering for review count does not narrow the field to better products. It narrows it to products that all look the same.
We hold 66,150 product listings captured across six Indian marketplaces — Amazon India, Flipkart, Nykaa, Nykaa Fashion, Croma and Reliance Digital. Of those, 30,470 display a star rating. We sorted them by how many reviews stand behind the rating, then asked one question: does thicker evidence produce a wider spread of scores?
| Minimum reviews | Listings | Rated 4.0★+ | Median rating |
|---|---|---|---|
| Any rating shown | 30,470 | 84.3% | 4.30 |
| 100 or more | 12,501 | 94.9% | 4.30 |
| 1,000 or more | 5,645 | 97.5% | 4.30 |
| 10,000 or more | 1,741 | 98.2% | 4.30 |
Read the last column first. The median rating does not move at all. Whether you look at everything or only at products with five figures of review volume, the middle of the pack sits at 4.30. What changes is that the low end disappears: at any review count, roughly one listing in six falls below 4.0; above 10,000 reviews, fewer than one in fifty does.
So the filter does something — just not what people think. It is not surfacing better products. It is removing the products that had a visible flaw, and leaving you with a crowd that is indistinguishable by rating alone.
Of the 25,689 listings rated 4.0 or above, 66.1% sit between 4.0 and 4.4. Two-thirds of the entire top of the market is packed into four-tenths of one star. Only 3.8% of everything rated sits below 3.0.
That is what a five-point scale looks like when almost nothing scores in its lower half. The gap between a 4.2 and a 4.5 is real arithmetic, but it is a sliver of a range that is barely being used — and it is doing far more work in a shopper’s head than it can support.
Something genuinely useful, just not what it gets used for. Review count tells you how much to trust a rating. It does not tell you how good the product is.
A 4.3 backed by 12,000 reviews is a dependable 4.3 — it is unlikely to move much, and it is not the product of six friends and a launch discount. A 4.3 backed by six reviews is not really a 4.3 at all; it is a guess with a decimal point. Both display identically on a marketplace listing page.
That is a confidence signal, and it is worth having. The mistake is reading it as a quality signal on top of the stars, when the data says it adds almost nothing there.
Every figure on this page comes from our study The 4-Star Rating Problem, an analysis of 30,470 rated listings captured across six Indian marketplaces. That page carries the full method, the platform and category breakdowns, three robustness checks — including one that tries to break the headline — and a CSV of every published figure you can download and check yourself.
What this is not: our capture set is not a random sample of Indian e-commerce, and our coverage is uneven. We also did not test whether any individual review is genuine, incentivised or bought — that is a different question and this data cannot answer it. Both limits are set out in full on the study page.
Often not meaningfully. Both sit inside the band that holds two-thirds of all high-rated listings. Treat a gap that small as noise unless something else — review depth, written complaints, price, specification — agrees with it.
No, and we want to be exact about this: we did not test it, and this data cannot show it. Compression like this has perfectly ordinary explanations — most obviously that people happy enough to complete a purchase are likelier to leave a rating than people who quietly returned the item. We are describing what the displayed numbers do, not accusing anyone of anything.
Yes, within the limits of what we can report. It appears in all three marketplaces and all nine categories whose samples are large enough to publish, measured separately. The per-platform and per-category tables are on the study page so you can check rather than take our word for it.
We publish two numbers instead of one: a quality score anchored to the displayed rating, and a separate confidence level set by how much evidence stands behind it. Where the evidence is too thin we print no score and say so, rather than inventing a number. Roughly seven in ten products in our catalogue currently fall into that category. The full formula is on our methodology page.
Snapshot: 19 August 2026, n = 30,470 rated listings. Figures regenerate from a single script — see how to reproduce this. Quote or reproduce with attribution and a link; no permission needed.