Only reviews move the board
A product enters a ranking the moment it has one real review, and its position comes from those reviews alone. Nobody — not a brand, not a retailer, not us — can buy a spot, nudge a row upward, or remove a competitor. When a brand pays for visibility it appears in a separate slot that is always labeled Sponsored, above and apart from the ranked table, and it never changes the organic order underneath.
The math, in plain language
The star rating you see on each row is the product's raw average — we never massage the displayed number. But boards are ordered by a dampened version of that average, because raw averages are easy to fool at low volume: one enthusiastic 5★ review would put a brand-new product above a favorite with hundreds of reviews.
The damping works like this: before counting a product's real reviews, we pretend it already has two invisible, perfectly average votes of 3.5★. In formula terms:
score = (average × reviews + 3.5 × 2) ÷ (reviews + 2)
A lone 5★ review scores (5 + 7) ÷ 3 = 4.0, while a 4.6 average across ten reviews scores about 4.42 — so the well-reviewed product ranks higher, honestly. As real reviews accumulate, the two invisible votes fade into irrelevance and the score converges on the true average. Ties break by review count, then by a stable id.
Small samples are flagged, not hidden
Damping softens the low-volume problem; it doesn't erase it. So any board built from fewer than 25 total reviews carries a visible Small sample — treat as early signal note. Early boards are real data, honestly ranked — they're just not yet a verdict, and we'd rather say so than let a thin board wear the authority of a deep one.
Lenses re-order, never re-invent
The Top rated, Trending, Most popular, and Best value views all rank the same reviewed set — they only change the sort. Trending uses recent view velocity, Popular uses 30-day views, and Best value divides the dampened score by price. In every lens the dampened rating stays the tiebreak, so a low-signal product can't leapfrog on a technicality.
Price per volume, only when it's real
The $/ml column divides a product's price by the volume parsed from its label size ("150ml", "1.7 fl oz"). When the size isn't a parsable volume — weights, sheet counts, multi-product kits — the column shows "—" rather than a guessed density. Same rule as everywhere else on beautydew labs: derived numbers only, never invented ones.
The Shop column is a convenience, not a ranking input
Some rows carry a Shop link to a real retailer offer — some of those are affiliate links, and we may earn a commission if you buy. That commission never affects a product's score or its position on the board: the ranking above is computed entirely from reviews, and a product with no offer at all ranks exactly where its reviews put it, Shop link or not.
The provenance stamp
Every board states when it last changed and how many reviews it stands on ("Updated {date} · ranked from {n} reviews"). A ranking built from reviews changes exactly when its reviews do — so that stamp is the board's honest freshness, not a marketing pulse.