Trust is a standard. Not a sticker.
Every offer starts with the same question: does this seller have enough credible history to be recommended?
01. First, the non-negotiables.
A seller must meet every hard gate. Missing, invalid or stale required evidence means the offer stays hidden.
02. Evidence earns the score.
We apply the 95% Wilson lower confidence bound to the positive rating. A tiny review sample therefore carries less confidence than thousands of reviews. Review and order volume use logarithmic scaling to reward evidence without letting volume overwhelm other factors.
Weighted factors: rating 30%, reviews 10%, dispute 16%, tenure 10%, activity 6%, orders 10%, delivery 10%, resolution 5%, protection 3%. Missing protection data earns a neutral 50 for that factor; we do not infer coverage. Seller metrics remain visible in each offer’s explanation.
We evaluate thresholds before rounding. Hard-gate failure always hides the seller, even if a score is high.
03. Then, find the right deal.
Best Overall combines Trust Score (65%), relative price (20%) and relative delivery (15%). Best Price and Fastest Delivery are selected exclusively from the trusted pool. Maximum estimated delivery time is used for delivery ranking.
Trust has limits. We show them.
A recommendation is not a guarantee. Scores reflect available marketplace data, can change, and do not verify every product claim. Always check the listing, seller terms, region and provider protection before buying.
Algorithm version: 1.0.0-wilson. Demo evidence is synthetic.