ORIGINAL SOURCE-DATA RESEARCH
DHgate spreadsheet photo audit: what 999 source rows showed
By DHgate Edit · Published 2 November 2026 · Source batch: 1 November 2026
Our DHgate spreadsheet photo audit examined image references attached to 999 affiliate-export rows. We could decode 998 row-linked references and found eight exact-repeat groups, with two references in each group. These counts describe image reuse. Seller and product equivalence require separate checks.
A copied photo is a reason to investigate a listing. It does not establish what a seller holds in stock, what an order will contain or whether two links lead to equivalent goods. This audit explains the evidence we have and the decisions it can support.
Results from this batch
The table describes one batch supplied by the site owner. It is not a random sample of DHgate, a catalogue-wide seller assessment or a quality rating. Some source images are collages containing multiple colours or designs.
| Metric | Count |
|---|---|
| Source rows in this batch | 999 |
| Readable row-linked image references | 998 |
| Image references not decoded | 1 |
| Distinct decoded RGB images | 990 |
| Exact-repeat image groups | 8 |
| References inside repeat groups | 16 |
| Repeat occurrences beyond the first | 8 |
Download the dated counts and definitions (CSV) ↓
“Eight repeat occurrences” counts the extra copy after keeping one reference per exact-repeat group. “Sixteen references inside repeat groups” counts both the first and the repeated occurrence. Neither figure means sixteen distinct products were tested.
The 990 distinct decoded images are also not a count of 990 unique products. One collage can show many products. The same product can appear in different pictures, crops or advertisements. Image uniqueness and product uniqueness answer different questions.
How we checked the source images
- We matched each returned batch item code to the corresponding export record and image reference.
- We downloaded the public image assets listed in those records. Two rows already known to share an image URL reused the cached asset instead of downloading it again.
- We attempted to decode each row-linked image. One downloaded file remained truncated after a retry, so it was held out rather than used for a product card.
- For readable images, we applied EXIF orientation handling, converted the image to RGB and compared its dimensions and decoded RGB pixel bytes using a SHA-256 fingerprint.
- We grouped exact matches, reviewed contact sheets and retained one candidate per identical-image group. We also held out broad style ranges that overlapped catalogue coverage when a distinct additional option was not established.
The process combines automated exact-image comparisons with AI-assisted visual review. It does not include test purchases or first-hand inspection of physical goods.
The pixel comparison runs on the supplied source images before catalogue cutouts or AI-prepared references are made. It is designed to find exact decoded-image reuse. The fingerprint does not inspect a seller profile, open an order or test a product.
A first-pass contact-sheet review was completed across the batch, with the unreadable image explicitly flagged. That first pass does not finish the closer variant, seller and current-listing checks needed for a recommendation. Our catalogue methodology separates those stages.
What this audit cannot prove
- Same image does not prove same goods. Two listing numbers can reuse a picture while differing in seller, option, construction, price or availability.
- Different image does not prove a different product. A crop, new watermark, resized copy or colour edit can change the pixels while showing a related item.
- A collage is not a confirmed option list. Its photographed designs may not all be available, and a shared affiliate link does not preselect the pictured colour.
- We did not audit every seller or place sample orders. The image review establishes no authenticity, materials, durability, delivery or commission-attribution success rate.
- This is a dated source-data snapshot. The export is not a live checkout quote. An image that failed to decode here is not proof of a permanently broken listing.
The underlying affiliate export is not republished in this download. The CSV contains aggregate counts and their definitions, so readers can see exactly what the summary measures.
What to do when two spreadsheet entries look the same
Start by recording both item codes. Open the current listings yourself and compare the seller identity, exact option, quantity, destination and delivered cost. Keep those facts separate from the picture. If one link opens a different item, stop and use the listing-link troubleshooting checklist.
Always ask for current QC photos of the exact colour, size and design the seller would supply. Request views that answer your actual uncertainty: front and back, fastenings, labels, measurements and close-up details. A familiar catalogue image does not replace that request.
Compare like-for-like costs after confirming the option. A cheaper exported price does not settle whether the seller supplies the same item. The price and shipping guide explains why unknown shipping must remain unknown.
How this changes the catalogue
We use the results to hold repeated-image candidates out of new imports. Product-name and pictured-option checks provide additional safeguards; manual comparisons are still necessary. When several distinct colours are pictured in one source listing, they can have separate cards with the same clearly labelled listing link.
Supplier-image cutouts retain the source product pixels. AI-prepared references are labelled and are not counted as independent product evidence. The original supplier image remains available in the details so you can compare the reference with its source.
Browse the pictured Batch 1 finds or return to the searchable DHgate spreadsheet catalogue. Treat both as research starting points, then confirm the current listing before ordering.