How the Dive Score works
Every number on this site comes from the product's public AliExpress listing. Here is exactly what we do with it.
1. We collect the listing data
When a product enters our queue, we fetch its listing: title, photos, current and list price, average star rating, number of reviews, number of orders, the store name and the seller's specifications. Live products are re-checked every week so prices and ratings don't go stale.
2. We calculate a score from 0 to 10
- Buyer rating — 60%. We don't take the star average at face value. It's blended with a neutral 4.3★ baseline in proportion to how many people reviewed it (a Bayesian average). A 5.0 from 4 reviews lands near 4.4; a 4.8 from 5,000 reviews stays at 4.8. Adjusted ratings from 3.8★ to 4.9★ map onto the full range.
- Order volume — 30%. Orders are measured on a log scale: going from 100 to 1,000 orders counts as much as going from 10,000 to 100,000. 100,000+ orders earns full marks. When a listing doesn't report its order count, this part is left out and the other two are re-weighted, rather than counted as zero.
- Markdown — 10%. The current price compared to the seller's own list price. 60% off or more earns full marks. It's weighted low on purpose: list prices on marketplaces are often inflated.
3. We label the result
-
9.0 Deep find
- 8.5 and up — Deep find. Excellent rating backed by a large number of buyers.
-
8.0 Great catch
- 7.5 to 8.4 — Great catch. Strong on most signals.
-
7.0 Solid pick
- 6.5 to 7.4 — Solid pick. Good, with a weaker signal somewhere (often fewer orders).
-
5.0 Worth a look
- Below 6.5 — Worth a look. Newer or less proven. Read recent reviews first.
4. What we write
The verdict, "What holds up" and "Watch out for" notes on each report are generated from the same data — nothing in them is invented. When an editor adds a personal note, it's shown separately above the generated summary.
What the score is not
We haven't physically tested these products, and the score can't tell you whether a specific size or color variant suits you. It tells you how the listing has performed for the people who already bought it. Commissions never change a score — the formula doesn't know which products pay more.