What we know, how we know it, and what stays yours.

Why this page exists

Most platforms don't tell you how product data is created, where it comes from, or who owns it. We think that's a problem, especially when automated systems are involved.

Patina Studio uses automated processing to analyse product photos, extract attributes, research brands, assess condition, and write descriptions. That's a lot of automated decision-making about something you're buying or selling. You deserve to know how it works.

How product data is created

Every listing on Patina Studio starts with photographs. A seller uploads images of their product, and Patina analyses them in four stages.

Classification. Each image is categorised by what it shows: the product itself, the brand label, the care label, construction details, or any defects. This determines how much evidence the system has to work with.

Extraction. Patina reads the classified images and pulls structured data: brand, materials, colour, sizing, closure type, pattern, and more. For condition, it assesses specific wear markers like pilling, fading, and marks.

Enrichment. Patina researches the brand, identifies the product line, and adds the context that makes a listing sell. It identifies the model, colourway, season, and estimated era where possible.

Composition. All extracted and enriched data is composed into a structured listing: a written description, a formatted title, a condition assessment, and marketplace-ready output.

At every stage, the seller reviews and approves the output before anything is published. Patina proposes. The seller decides.

What Patina does and doesn't do

Patina does: identify brands and products from label photos, extract material composition from care labels, assess visible condition from product images, research brand context from known sources, write product descriptions from structured data, and suggest a condition grade based on visible evidence.

Patina does not: authenticate items (we flag indicators but do not guarantee authenticity), set prices (pricing is always the seller's decision), auto-publish listings without seller approval, make up information it cannot extract from images or known references, or use your data to train external systems. Does not set prices or auto-publish. You approve every listing before it goes live.

If Patina cannot determine an attribute from the available evidence, it leaves the field empty rather than guessing. An empty field is more honest than an invented one.

Where the data lives

Product data is stored in a structured database. Every attribute, from brand name to heel height, is stored as a discrete, typed field, not as free text buried in a description paragraph.

This means the data that powers a listing is the same data displayed in the product detail page, the same data used for search and filtering, and the same data that generates the marketplace output. There is one source of truth per product, not multiple conflicting versions across different platforms.

When a product is listed on multiple marketplaces (Vinted, eBay, the Patina marketplace), the underlying data is identical. Only the formatting changes to suit each platform's conventions.

Data quality and accuracy

We take accuracy seriously, but we are honest about limits.

Extraction is not perfect. It works best with clear, well-lit photographs of readable labels. A blurry care label or an obscured brand tag will produce less reliable results. We score the evidence quality of every listing so sellers know when their photos need improvement.

Brand research is sourced from known references, not invented. If Patina cannot find reliable information about a brand or product line, it doesn't fabricate detail. The description will focus on what's visible and verifiable.

Condition grading is based on visible evidence in photographs. It is a structured assessment, not a guarantee. We encourage sellers to review condition grades and adjust where their hands-on knowledge exceeds what Patina can see in a photo.

Every attribute is editable. If Patina gets something wrong, the seller corrects it before publishing. Published data reflects the seller's approval, not just Patina's output.

Who owns what

Your product data is yours. Images you upload, descriptions generated from your products, and attributes extracted from your photographs belong to you. You can export your data at any time.

Your products, your data, your photos. Listing on the Patina marketplace doesn't transfer ownership of anything. If you leave Patina, your data leaves with you.

By using Patina Studio, you grant us a licence to display your product data on the Patina marketplace and to format it for connected marketplaces. This licence exists so we can operate the platform. It does not transfer ownership.

If you delete your account, your data is removed within 30 days, except where items are part of active transactions or where we are legally required to retain records.

Processing and third-party services

Product images and data are processed by Patina's enrichment systems to power extraction, enrichment, and description generation. Your data is never used to train external systems. What Patina learns from your products stays within Patina.

We transmit the minimum data necessary for each processing step. Images are sent for visual analysis. Text is sent for language generation. We do not send your personal information (name, email, payment details) to processing services.

Our enrichment infrastructure may change over time as we improve our capabilities. We will update this page to reflect any material changes in how your data is processed.

Transparency in listings

Every listing on the Patina marketplace shows structured, verifiable data. Buyers can see exactly what was assessed: the condition grade and the specific detail fields behind it, the materials and construction, the provenance and origin.

We don't hide behind vague descriptions. If a listing says "Pilling: None", that's a specific, disputable claim. If a buyer receives an item with pilling, the listing data is the reference point for resolution.

This structured approach means buyers don't need to trust the seller's writing ability. They trust the data, and the data is governed by the platform.