Verity / Entrupy technology
Entrupy is a machine authentication system for luxury goods, built on microscopic imaging and computer vision. It answers the question that carries the highest price tag in resale: is this the original, or a high-grade replica? Below — how the analysis works, what lies behind the stated accuracy, and why the verdict of an algorithm is not the same thing as the opinion of an expert.
Entrupy Inc. is a US technology company founded in 2014. The technology grew out of research at New York University: four years of work preceded the product, and at its core are patented computer vision algorithms and a method of microscopic imaging of materials. One of the co-founders is NYU researcher Lakshminarayanan Subramanian; the company is led by Vidyuth Srinivasan.
The solution has been recognised with an Edison Award and an AI Breakthrough award. Among the clients Entrupy names publicly are StockX, the sneaker and streetwear marketplace, the US pawn shop chain Max Pawn, the consignment chain Clothes Mentor and Goodwill.
A check begins with photography. The item is photographed with the smartphone's standard camera at high magnification, from a fixed set of angles: the app itself indicates which area to capture and from what angle. The set of frames is not arbitrary — it is a protocol, identical for every check of the same type of item, and it is precisely this repeatability that makes results comparable with one another.
Then the model takes over. The captured images are compared against a database holding millions of records of genuine and counterfeit examples, and the algorithm analyses millions of data points on every check. What it looks at is what cannot be reproduced in a backstreet workshop — or even under factory conditions:
The result comes back within minutes. For certain categories of especially expensive items the analysis may take up to twenty-four hours.
There is no separate device for running a check. A few years ago an optical magnifying attachment was fitted over the phone camera, and someone had to bring it along; modern cameras manage without it. Throughout all that time the analysis was never performed inside a device but on a server — the attachment only helped obtain the required magnification.
The objection comes up almost every time: how can authenticity be judged from images if the specialist never held the item? The answer is that handling the item in person is exactly what tells you the least today.
Copies at the top of the market are made from leather of comparable quality, with the same hardware and a stitch pitch close to the original. A top-price-segment copy is indistinguishable from the genuine article by touch, by smell and by weight — and an experienced buyer comparing it against a remembered impression gets it wrong systematically, not through carelessness. The discrepancies do not disappear, though: they simply move to a scale the human eye physically cannot resolve. In that sense a magnified image is more informative than an inspection in person, and comparison against a reference database is more informative than comparison against a memory.
The figure is public — the developer itself publishes it on its website — and it refers to the technology as a whole, to the share of correct conclusions across a sample of checks. It is not a guarantee for an individual item and not a promise anyone makes to the owner of a particular bag: the correct wording is "verification accuracy as stated by the developer".
The practical meaning of the figure lies not in the nines themselves but in the nature of the error. An algorithm's error is random and the same for everyone: it cannot be induced deliberately, cannot be tilted in anyone's favour, and cannot be explained by a specialist being tired that day or having a stake in the outcome of the deal.
A classic appraisal is the judgement of one particular person. It cannot be reproduced: a second specialist is entitled to reach a different conclusion, and any opinion remains open to dispute between buyer and seller. That is precisely why an argument about authenticity so rarely reaches an end: it rests on authority rather than on a verifiable fact.
An Entrupy result is reproducible. It is produced by an algorithm, it does not depend on who photographed the item, and it returns the same conclusion from the same images. For a pawn shop, a consignment store or a resale platform, that is where the real value lies: the conclusion stops being an opinion and becomes a document that can be cited to a client, to an insurer, or in correspondence over a disputed transaction.
A confirmed check produces a digital certificate: the date, the item's details and a QR code by which the authenticity of the certificate itself can be verified at any moment. Entrupy certificates are accepted by resale platforms worldwide and are used across the market as a standard form of confirmation on resale.
What the certificate looks like, what it confirms and how to verify it by QR code →
Every certificate issued is backed by an Entrupy financial guarantee valid for one year from the date of issue; the company publishes the current terms on its own website. If the authenticity of an item is subsequently disputed, the item is re-examined from scratch — including by independent experts — and it is Entrupy that carries the obligation under the guarantee, not the seller and not Verity.
That is the practical difference between a certificate and an opinion: behind the first stands a company with publicly published obligations and a certificate number anyone can check, behind the second stands the reputation of one individual, which the other side of the deal can neither verify nor hold to account.
Categories: handbags, backpacks, wallets, belts and leather accessories. Brands — more than twenty: Hermès (including Birkin, Kelly, Constance), Chanel, Dior, Goyard, Louis Vuitton, Gucci, Prada, Saint Laurent, Balenciaga, Bottega Veneta, Burberry, Celine, Chloé, Christian Louboutin, Dolce & Gabbana, Fendi, Givenchy, Loewe, Miu Miu, Salvatore Ferragamo, Valentino, Coach, MCM.
The list is set not by a brand's popularity but by the availability of a sufficient reference base: a model becomes checkable once enough genuine examples have been accumulated for comparison. The base grows with every new check.
The flip side of the same rule: a brand that is not in the base cannot be checked. The algorithm has nothing to compare against, and any answer would be a guess — so no such check is carried out and none is charged for.
Authentication answers exactly one question — is the item genuine or not. It does not assess condition, does not establish the year of manufacture and does not confirm the completeness of the set: a single model can exist in several production variants, and a discrepancy with the picture on the brand's website means nothing in itself.
And no system offers a hundred per cent guarantee — it is more honest to say so plainly. The value of the result lies elsewhere: it rests on measurable features rather than on anyone's opinion, it is reproducible from the same images, and it can be put in front of the other side of the deal.
Verity applies Entrupy technology in Russia and provides authentication services to businesses — pawn shops, consignment stores and resale platforms — under a service agreement. The item stays with the partner: there is nothing to transport or ship anywhere. The history of every check, along with statuses and certificates, is stored in the Verity app and available to the business as reporting under the agreement.
Entrupy is not affiliated with the brands listed on this page and is not their representative. Authentication is based solely on Entrupy's algorithm; the reference database is not compiled by the manufacturers. The brands bear no responsibility for the results of Entrupy checks and may not recognise the certificates issued.
FAQ
It is a machine authentication system for luxury goods: the item is photographed at high magnification, and the algorithm compares the microstructure of the material, the stitching and the hardware against a database of genuine examples of the same brand. The conclusion is drawn by a trained model rather than by a person, so the result is reproducible and does not depend on who looked at the item.
The US company Entrupy Inc., founded in 2014 on the basis of research at New York University. Its computer vision algorithms and its method of microscopic imaging are patented.
The developer states a verification accuracy of 99.86%. This is Entrupy's public figure, published on the company's website; it refers to the technology and is not a guarantee for any individual item.
No. An optical magnifying attachment used to be fitted over the phone camera; it is no longer required — the images are captured with the smartphone's standard camera, following the app's prompts.
More than twenty leather goods brands: Hermès, Chanel, Dior, Goyard, Louis Vuitton, Gucci, Prada, Saint Laurent, Balenciaga, Bottega Veneta, Burberry, Celine, Chloé, Christian Louboutin, Dolce & Gabbana, Fendi, Givenchy, Loewe, Miu Miu, Salvatore Ferragamo, Valentino, Coach, MCM.
Only that Entrupy does not hold a sufficient set of references for that particular model to compare against. It is neither a finding of counterfeit nor a finding of authenticity: the check technically cannot be performed, and the service is not charged for.
Yes. Verity applies Entrupy technology in Russia and provides authentication services to businesses — pawn shops, consignment stores and resale platforms — under a service agreement.
Yes. Every certificate is backed by an Entrupy financial guarantee valid for one year from the date of issue; the company publishes the terms on its own website. If authenticity is disputed, the item is re-examined from scratch, including by independent experts, and the obligation sits with Entrupy — not with the seller of the item and not with Verity.
No. Authentication answers the question of authenticity only. It does not establish condition, year of manufacture or the completeness of the set: a single model can exist in several production variants, so a discrepancy with the image on the brand's website means nothing in itself.