One of the leading fashion marketplaces with a catalog of 650,000+ SKUs and a flow of 100,000+ new arrivals per month checks size charts against brand official websites. Manual verification of a single SKU took 8–12 minutes — a browser agent scaled the task to the entire catalog.
Category managers and the content team are responsible for the accuracy of all product attributes, including size charts — a key parameter that directly affects returns, conversion, and customer satisfaction.
Spot checks confirmed the hypothesis: size charts on listings frequently diverge from what the brand specifies on its own website. Suppliers provide data in different formats, errors accumulate over years, and manual verification at catalog scale is physically impossible.
Discrepancies lead directly to returns. In the pilot sample, a top brand's size M on the platform was listed as "bust circumference 88–92 cm," while the brand's website showed "90–94 cm"; a global footwear brand was missing foot-length fields that were present on its website. Each case is a potential return and negative experience.
Manual verification of a single SKU took 8–12 minutes: find the website, find the listing by supplier article number, open the table, compare dozens of fields. For 150,000 SKUs, that is approximately 25,000 person-hours. There was no systematic verification — responses were reactive, based on complaints.
A browser agent cross-checks size charts in product listings against official brand websites: auto-search, Vision-based table extraction, and size system normalization.
Using the brand name and supplier article number, the agent queries the Serper Google API and, based on signals (domain, content, catalog), autonomously selects the official website, caching the result — one brand = one search for the entire batch. Most brands are matched automatically; the rest are re-verified on Wildberries with the source flagged.
Built on browser-use Cloud with proxy rotation, the agent behaves like a real user: opens a listing, clicks a size, expands the table, scrolls. Parsing does not depend on the specific frontend of the site — even a CMS change does not break verification, and proxies provide access to international sites.
The async FastAPI + asyncio architecture launches thousands of virtual browsers in parallel; grouping by short_sku and caching by brand reduce the operation count by orders of magnitude. Scheduled overnight runs deliver the report by morning.
When a brand website is inaccessible, the agent falls back to Wildberries, where size charts are often published as images. Claude Vision extracts measurements, sizes, and units from a screenshot with the same accuracy as from HTML — the non-standard case is handled completely.
A React interface with drag-and-drop Excel upload, real-time progress via Server-Sent Events, color-coded status (ok / error / not found), filters, and .xlsx export with source links. Category managers work independently without IT involvement.
Size charts transformed from a "blind spot" into a verifiable catalog asset.
The agent framework is reusable beyond the original task — product attribute verification, competitor price monitoring, UGC collection, working with any source without an API. The solution becomes a platform of the class "AI agent operates on the web like a user."
Previously, we saw size discrepancies through returns — months after a listing went live. Now we see them on the day the product arrives. The category team uses the tool on its own, without IT, on a daily basis.
The metrics and results presented reflect outcomes of a specific project and depend on its initial conditions. They are provided for informational purposes only, do not constitute a public offer, and do not guarantee similar results in other projects. Supporting materials are available on request.
Generative AI rebuilds SEO content and visuals for product listings: copy, hero images, and rich content tuned to marketplace algorithms — all in a single pipeline.
An AI pipeline assembles lifestyle videos from product photos: decomposition into short artifact-free segments, parallel generation, human in the loop.
An AI platform mass-processes product photos to a unified standard: background removal, studio shadow, cropping, and batch processing of thousands of SKUs.
No cases match the selected filters.
Tell us about your process — we will suggest an audit, pilot, or full deployment.