Findings · DOC-0009
What the data says
Findings across every store on record — 1,139 of them, 171,089 products — recomputed daily from what those stores publish about themselves. Scores run 26 to 85 with a median of 58; the top tenth starts at 73.
Every figure carries its own denominator, because “31% of stores” and “31% of stores whose shipping policy we could parse” are different claims and only one of them is true. Where a number invites a conclusion, the conclusion is labelled as a reading rather than mixed into the fact. Measured 2026-07-29, rules v0.8.
281 stores state a delivery window of 10 days or longer in their own shipping policy
Reading — A long window written by the store itself is the most-cited signal of overseas fulfilment. It is not proof of anything — plenty of legitimate makers ship slowly — which is why the number is stated with the policy sentence it came from rather than as a verdict.
17 stores mark down at least 70% of their catalog at a median 40% off
Reading — A permanent, catalog-wide sale is how an inflated "was" price looks from the outside. A real seasonal sale ends; this is measured every day, so a store that appears here is not having a weekend.
240 stores publish an affiliate program with a join link that works
Reading — Every creator-side catalogue of these sits behind a login. This is the same information, read from the brands’ own pages.
490 stores publish a free-shipping threshold or free shipping on every order
Reading — Newest of the signals here, so the denominator is small until every policy has been re-read on the nightly pass. It is stated against what we have actually looked at rather than against the whole registry, because the alternative would report our collection progress as a fact about stores.
21 stores have a catalog that is at least 90% out of stock
Reading — A storefront that is still up while almost nothing can be bought — usually paused or abandoned rather than dishonest, but worth knowing before you send an audience to it.
1 stores appear on a malware or phishing blocklist
Reading — In every check so far the listed domains have been established brands with an apparently compromised site, not the ones that look sketchy. The score is never docked for it: a blocklist entry can be stale or wrong, so it is published as a flag with its source instead.
Using these numbers
They are free to quote under CC BY 4.0 with a link. Quote the denominator with the figure — without it the sentence says something we did not measure. Every number here moves daily as stores change what they publish, so cite the date as well; the dataset page has the ready wording, and the changelog says what each formula version changed.