How we score listings

The Dhundo listing score answers one question: is this a good listing to buy from? It is our own assessment, built from things we can observe about the listing itself.

It is not a customer rating. Nobody has voted on it, and we do not publish it as though they had. It is also not a measure of how well a product matches what you typed - that is a separate number, and the two disagree often enough that showing one under the other's name would be misleading.

We do not claim a listing at one store is the same item as a listing at another. Nothing in the data we can see supports that, so we do not say it. Where we show listings from other shops, we call them similar listings, because that is what they are.

Five signals, weighted. A listing needs at least 3 of them before we will publish a score at all - below that we say we have not scored it rather than guessing.

Rating strength

How well this is rated, adjusted so a lone 5-star review does not outrank a product with hundreds of good ones.

How many people rated it

More ratings means more confidence in the rating.

Popular for its store

How this compares to other products on the same store, so a busy marketplace does not automatically beat a small shop.

Recently seen

How recently we confirmed this listing is still live and priced.

Listing quality

Whether the listing itself is clear and complete - a real title, a usable image, enough detail to know what you are buying.

This is not the score above - it is a separate check on the price itself. Every time someone clicks through to a store, we record the price our card showed them. We can then compare that against our own data to see whether it was accurate.

Over the last 30 days, 81.1% of 37 click-outs we could check showed the same price we had most recently read for that listing. The other 7 showed a price that had since moved in our own data - our card was citing a superseded read, which is our defect. 162 more click-outs in the window could not be checked at all (no matching price read on our side) and are not counted in that percentage.

A related but different question: did the price still hold once the shopper actually arrived? Comparing against the next price we read after the click, 71.4% of 14 adjudicable click-outs still matched. A mismatch here is not necessarily our fault - the store may have changed its own price in that window - but it is still the shopper's experience, so we report it on its own rather than folding it into the figure above.

No affiliate postback, retailer webhook or any other observation of a purchase exists, so a conversion rate here would be invented. This section is about whether our own number was right, not about what happened after - we cannot see that, so we do not claim to.

Paying us cannot change a score or a position. Sponsored placement is a separate strip, labelled where it appears, and it is not an input to this calculation - the code that computes the score is not told whether a listing is sponsored.

A store cannot raise its own score by giving us more data, and we do not accept payment to review, re-rank or remove a listing.