How to Read Product Ratings on Amazon, Flipkart and Myntra (2026)

Over two days in September 2026 we read 360 listings across six shelves on five Indian shopping platforms - bottles, shoes, bags, keyboards, routers - and recorded the rating and the rating count each listing displayed.

The averages looked broadly similar everywhere: most listings sit between 3.8 and 4.6 whatever you are buying and wherever you are buying it. The counts did not. The median listing on Flipkart's bottle shelf carried 3,155 ratings. The median listing on Tata CLiQ's running shoe shelf carried eleven. Those two numbers are displayed in exactly the same way, as a star figure with a small number beside it, and they are not the same kind of evidence at all.

This page is about what that small number is actually counting on each platform, because it is not the same thing on any two of them.

Every platform here has to answer the same question: a product comes in six colours, so is that one listing or six? The answer determines what a rating count can possibly mean.

Amazon and Flipkart pool. One product's variations share a review pool, so all six colours show the same count. In our Amazon read, three United Colors of Benetton bottles at the identical Rs. 349 all displayed 173 ratings. On Flipkart, a Prestlee bottle listed at Rs. 245, Rs. 699 and Rs. 992 displayed the same 3,155 ratings against all three prices.

Myntra pools wider. Its counts attach to a style family rather than a product, which is how two different HIGHLANDER shirts at Rs. 495 both came back showing 227,002 ratings in our 17 September read, and two Bewakoof slings in black and olive both showed exactly 171.

Tata CLiQ and Croma split. Each colourway is its own listing with its own rating field, and on a smaller platform those fields mostly stay empty. The Asian Wonder 23 running shoe appeared three times on Tata CLiQ at Rs. 604 with no rating on any of them. Logitech's Pebble 2 keyboard combo appeared four times on Croma at Rs. 4,595 with three ratings between the four listings - one showing 5.0 from a single rating, one showing 3.0 from two, and two showing nothing.

Because the mechanics differ, the threshold at which a count becomes useful differs too - and it is much lower on the smaller platforms than shoppers assume.

On Amazon and Flipkart, four-figure counts are ordinary. Our two bottle shelves had median counts of 2,817 and 3,155, so a three-figure count there is a below-average listing rather than a well-evidenced one. On Myntra it depends on the aisle: running shoes had a median of 1,492, but its bag shelf had a median of 170 and twelve listings with nothing at all. On Croma the median was 51, and on Tata CLiQ eleven - so on those two, a count in the low hundreds is genuinely strong evidence.

The floor is the same everywhere, though. Below about twenty ratings, an average is noise: one unhappy buyer out of five moves the figure by a full star, and a 5.0 from a single rating - which is what one Croma listing showed us - tells you precisely one person's opinion while occupying the same space on the page as a verdict from fifteen hundred.

None of this requires any special tools. It is five habits, and they take about thirty seconds each.

The fifth one is the least obvious and the most valuable. On a T-shirt, a bad product announces itself on day one and a hundred ratings will tell you. On a vacuum flask, a folding keyboard, a wheeled suitcase or a pair of running shoes, the failure is a seal that stops sealing or a hinge that stops hinging, eight months in. That is exactly what a six-figure count captures and a specification list cannot, which is why we lean on rating counts hardest on the products with the slowest failure modes.

And one thing to stop doing entirely: filtering a category by star rating. On every shelf we read, that filter promotes listings with three ratings above listings with three thousand, and on the split catalogues it hides the best-evidenced version of a product behind an empty duplicate of the same thing.