QC data: sources, counting rules and limits

QC photos on QCRadar are inspection photos taken in agent warehouses when a real order arrived for checking. We aggregate them from multiple agent platforms' inspection records and never intentionally remove the warehouse's own marks: original watermark, printed date and QR code where the platform prints one. Public copies carry a small QCRadar hosting mark and site QR in the corner — it marks the copy we host and distribute, and it is not proof of warehouse origin. We never fill a gallery with look-alike photos of the same style.

Photos submitted by signed-in buyers appear in product galleries labelled as user submissions after review, and are counted separately in our statistics.

One QC photo = one image file from an inspection record. A single inspection usually produces several photos (angles, measurements, defects), so photo counts run higher than order counts. Videos are excluded from photo counts. Duplicate files are dropped on ingest by image hash, so receiving the same file twice never inflates a number.

Photos are stored per source listing (platform + item ID) — what we call a SPU. Colour variants of the same listing share that pool, and each photo is assigned to its variant by the colour label on the inspection record. All aggregate statistics deduplicate at SPU level: a listing with twelve colours counts once, not twelve times. Where several sellers offer the same product, card badges show the best-documented seller’s count.

Total QC photos = all inspection photos in the catalog. “Products with QC” and catalog coverage count active products whose SPU has at least one photo. Category and brand rankings sum SPU-deduplicated counts; a collaboration item counts once for each member brand. All index numbers are recomputed as a daily snapshot — the “data as of” date on the page is the snapshot time.

Two of these numbers are counted at different levels, on purpose. The product count on the home page counts accessible product pages — one per colour variant, each with its own page — because it answers “how much is there to browse here”. Every QC figure instead deduplicates to the SPU, as described above. So one source listing with twelve colourways contributes twelve to the product count and one to the QC figures.

Inspection dates are parsed from the inspection record itself — the warehouse’s file path or filename. When no date can be parsed, we show nothing: we never substitute our own import time for an inspection date, because import time reflects our crawling schedule, not market activity. “Recently inspected” lists sort by these parsed dates.

One scope note: the “recently inspected” wall on the QC page ranks fresh inspection dates within the catalog’s popular pool, not across every product we track. The index numbers above it are catalog-wide.

We don’t rank agent platforms by QC volume. Our per-platform counts reflect how deeply we integrate with each source, not how popular a platform is — publishing that as a market ranking would be misleading, so we don’t.

We’ve tracked QC sites in this hobby for years. Judging whether a site really holds QC data doesn’t require trusting its homepage — and you should never judge by the showcase items it chose to display. The direct test is to pick a random, real product link from Weidian, Taobao or 1688, paste it into the site’s QC finder, and check whether it returns the matching product, the matching shop, and a traceable inspection record.

A real QC record forms a chain you can audit: the displayed product matches the final source link, photos keep an identifiable inspection source, a date or inspection details are visible, and the same SKU accumulates different dated batches as real orders come in. Not every agent prints QR codes, weight and size on the photo — but a site that can’t take a random link, and can’t show which product, which shop and which inspection a photo belongs to, is a QC picture gallery, not a verifiable QC finder.

Anyone can reproduce this: don’t search only the items a site promotes — test a handful of real links from different shops and platforms, and audit each result against its final source link. After years of re-checking the industry’s popular sites, the ones that reliably pass are few and far between.