Two errors in twenty products — and why nobody would have noticed
We ran a real twenty-product catalogue through our own listing check — not lines invented to fit our own parser, but live descriptions from a working store. Below is what it found, including what it found in our own code.
The numbers
| Measured | Result |
|---|---|
| Products in the run | 20 |
| Semantic errors in finished listings | 2 |
| Of those, crashed or raised a warning | 0 |
| Lines with no measurable specs | 17 of 20 |
| Run time | under a second |
Error one: a lip stain became a thermos
“A lip stain that will not rub off, even on a mug. 24 hours of wear”
“Keeps temperature for up to 24 h”
A number of hours was taken for heat retention. The listing did not crash, did not turn red and looked perfectly normal — it simply described a lip product as a vacuum flask. In a thousand-item catalogue this goes live and is found by the buyer.
Error two: a face cream became a leather item
“Skin like a filter. Skin as if filtered” (a skincare set)
“Material: leather”
Human skin taken for the product material. Same class of defect: quiet, plausible and completely wrong.
Why this matters more than it looks
What we did about it
Both are fixed by rules and locked down by tests we deliberately break — to make sure they actually catch something rather than merely being green:
- hours count as heat retention only if temperature is mentioned in words;
- material means genuine leather — not every mention of skin.
After the fixes, the same catalogue: 20 listings out of 20 with no findings.
An honest limit
Try it on your own product
Paste a supplier line — see the listing and the check right in your browser, no signup.
Open the checker Discuss a pilot