How AsliVerify verifies beauty products
AsliVerify runs a paired home-and-counter image check that fingerprints a product before it ships and re-fingerprints it at the return desk, then scores the two against the brand’s own SKU ruleset. The result is a tamper-evident product authentication record that closes the counterfeit detection gap beauty retailers have carried since online resale and parallel import made beauty counterfeits a category-defining shrink problem.
The seven-step flow.
Each step runs in sequence: the home capture lands, the counter rescan lands, the fingerprint comparison and brand rule lookup decide pass or flag, and the audit record arrives before the manager’s queue is touched. Pure explainer, no data plane.
- 01Step
Image capture at home and at the counter.
The shopper photographs the product from two fixed poses in the customer app, then the counter associate reshoots it from the same angles when the item arrives at the return desk. Two photos in two locations give the verify pipeline an honest before-and-after pair to work with.
- 02Step
On-device metadata extraction.
The customer’s phone extracts a structural fingerprint of the photo — packaging geometry, label-print vector signature, ink density profile — and hashes it before anything leaves the device. The full image never traverses the network during this stage, which is the first line of the privacy boundary we hold for product authentication work.
- 03Step
Brand rule lookup.
The fingerprint is matched to a brand-specific ruleset tied to that SKU: approved label revisions, registered distribution geography, expected packaging batch characteristics, and the known counterfeit signatures for that product line. Beauty counterfeits tend to repeat specific failure modes — wrong UPC weight, misspelled ingredient panels — and the ruleset encodes each one.
- 04Step
Similarity scoring against the home and counter pair.
The pipeline compares the home-scan fingerprint to the counter-rescan fingerprint, and scores each against the brand ruleset. For counterfeit detection this is where the home photo’s characteristics get cross-checked with what arrived at the counter: a high disagreement score is the strongest signal a product changed hands en route.
- 05Step
Pass or flag verdict.
The scoring pipeline returns a pass or a flag. A pass means the fingerprints agreed across both scans and the brand ruleset cleared the SKU; a flag means at least one rule failed or the two scans disagreed on a structural feature the brand has previously seen on counterfeit stock.
- 06Step
Tamper-evident audit trail.
Every verdict — pass or flag — is written into a serialized audit record with the SKU, customer id, scan timestamp, fingerprint hash, and verdict. The raw photos stay on-device through processing, and only the metadata fingerprint is uploaded for comparison, so the audit record is what the chargeback-defense team has later if the dispute opens.
- 07Step
Manager review escalation.
Any flagged verdict routes to the on-shift manager’s queue with the audit record attached, the brand-rule violation surfaced as a one-line reason, and the home-vs-counter fingerprint comparison waiting in the sidebar. The manager either releases the refund against the evidence or escalates upstream to LP for a representment package.
What gets captured.
Raw photos stay on-device through processing — only the metadata fingerprint is uploaded for comparison. Customer photos are never used to train third-party AI models. The seven fields below are everything the server sees per verification record.
Home-scan record
Metadata fingerprint from the shopper’s on-device capture, hashed before transmission.
Counter-scan record
Metadata fingerprint from the counter associate’s rescan at the return desk.
Fingerprint hash
A short, irreversible hash of the metadata fingerprint used for the similarity comparison.
SKU identifier
The product SKU tied to the verification record so the result binds to the right return.
Pass or flag verdict
The structured outcome the scoring pipeline emitted against the brand ruleset.
Timestamp
The scan moment in the customer’s and the counter’s local time, retained for the audit window.
Manager identifier
The on-shift manager attached to any escalated verdict, so the review chain is traceable.
Want to see the verdict arrive in the dashboard? Try the live demo — no signup, just a photo and a counter rescan.