Built closet.ai: one catalogue for 40 fashion stores (Lebanese and regional), re-read every morning — 38,000+ items with price history. The hard part wasn't the scraping, it was keeping it honest. Every store structures sizes, discounts and "sold out" differently, so the pipeline normalizes units and tracks each item's price over time instead of just re-listing it. That's how you tell a real sale from a mislabeled price, and catch when something actually goes back in stock. If you run a store and want to see how discoverable it actually is on Google, Maps or ChatGPT, I built a free check for that too: https://contra.com/s/RqlYXPWs-free-visibility-audit-and-indexing-fix