Gather: AliExpress review vetting sample (collected 27 September 2026)

WHAT THIS IS
Ten product types named as "winning" or example dropshipping products on vendor pages that rank for
"how to find winning products for dropshipping" (Zendrop's winning-products page and CJdropshipping's
"How to Find Winning Dropshipping Products"). For each type we picked one public AliExpress listing and
read its public reviews.

HOW THE LISTINGS WERE PICKED (sample-selection.csv)
For each product type we checked several AliExpress listings for that type and kept the one with the most
reviews (minimum 25). The posture corrector came from AliExpress search sorted by orders; the first result
there was a seat cushion, so we took the first actual posture corrector. This is a convenience sample of
ten listings, not a random or representative sample of AliExpress or of dropshipping.
Two types are near matches, not the exact product the vendor named: Zendrop's page lists an interactive
pet feeder toy, and the closest AliExpress listing we used is a motorised automatic cat toy with no feeder;
CJdropshipping names a foldable fan, and we used a portable neck fan as an adjacent form factor.

HOW REVIEWS WERE COLLECTED
On 27 Sep 2026 we collected all public reviews the listing showed from the 12 months to 27 September
2026: 3,946 in total across the ten listings, de-duplicated by review id, plus separate pulls filtered to
1, 2 and 3 stars. The feed showed nothing older than 12 months. "reviews_listed_by_endpoint" is the total
the feed reported;
"reviews_collected" is what we could actually read (a few percent fewer on most listings).
"reviewers_in_us" counts collected reviews (de-duplicated by review
id) whose reviewer country field was US; across the ten listings that is 302 of 3,946. The neck fan
listing's lifetime rating counts are higher than the 50 dated reviews the feed returned; every figure here
uses the dated last-12-months reviews.
Reviewer names were not collected. Review texts are not republished here.

HOW 1-3 STAR REVIEWS WERE CODED (low-star-reviews-coded.csv)
Every 1, 2 and 3 star review (290) was read in English (AliExpress machine translation where the original
was another language) and given one or more theme codes:
  FIT_SIZE                too small/large/short, does not fit the body, car or accessory it is meant for
  COMFORT                 hurts, digs in, chafes, too hard
  WEAK_OR_INEFFECTIVE     does not do the job it is sold for, weak power/magnet, battery, pet not interested
  CHEAP_BUILD             flimsy, thin, low-quality material or finish (without saying it broke)
  DAMAGED_OR_DOA          arrived broken or stopped working / came apart after little use
  SAFETY_DOUBT            reviewer doubts it is safe for its stated purpose (e.g. would not hold in a crash)
  WRONG_OR_MISSING        wrong item/colour, missing parts or pieces
  SHIPPING                never arrived, lost, very late, partial delivery
  NO_SPECIFIC_COMPLAINT   text present but no identifiable problem (often positive, e.g. "Good")
  NO_TEXT                 no text, or a single letter/word with no content
A keyword pass proposed codes; every review was then checked by hand and ~40 codes were corrected.
"low_star_product_defect_reviews" counts 1-3 star reviews carrying at least one of FIT_SIZE, COMFORT,
WEAK_OR_INEFFECTIVE, CHEAP_BUILD, DAMAGED_OR_DOA or SAFETY_DOUBT. "product_defect_per_100_reviews"
divides that by reviews_collected.

LIMITS
Ten listings. One coder. Machine translation. Star ratings and review visibility are controlled by
AliExpress. A clean review record does not prove a product is good; it is one input to a decision.