Mango Scraper for Catalogue, Price and Size Level Data

The eight digit reference, published on the product page as schema.org mpn. On its own it identifies a design, not a buyable item. The full key is reference plus colorId plus size id. Listing pages already use a composite tile id in exactly that shape, written as productId:colorId or productId:colorId:lookId.

One warning: the canonical tag is not a colour key. It normalises the look segment to 00, and on some references it points at the default colour rather than the colour you requested. Take colorId from the request path or from the payload.

Yes. Product URLs follow /(country)/(language)/p/(line)/(family)/(subfamily)/(slug)/(productId)/(colorId)/(lookId), and only the last three segments carry meaning. Request a reference with the wrong slug and the site answers 308 with the correct localised path, so we can crawl by reference alone and let Mango tell us the local name, category wording and canonical URL.

What differs per country is price, discount rate, promotion name, currency formatting, measurement units, warehouse ids and whether thirty day price history is present at all.

Not as counts. Mango exposes a boolean available per size, a lastUnits boolean that drives the visible "Últimas unidades" label, meaning last few items, on Spanish storefronts, an isDelayed flag, and a warehouses array of ids such as 400 or 001. Those ids are country specific, so they only mean something read against the storefront's own warehouse configuration.

Physical store stock is not published in the page payload, so we do not offer it. What we can do is sample the same size rows repeatedly and give you a sell-out timeline from the transitions.

Category URLs are /(country)/(language)/c/(line)/(slug)/(id), where the id is eight hex characters. The slug is not unique - the same word can appear under two different ids - so the hex id is the key we store.

The first server response carries 100 product tiles plus a pagination object with currentPage, totalPages, totalItems and hasNextPage. Everything after tile 100 loads client side. Faceted and sorted variants are disallowed in robots.txt, so we enumerate through category and subcategory ids instead of through filters.

Mango publishes no product or price API for outside teams, so the storefront itself is the read path: product and listing values arrive as escaped JSON inside the first server response, and that is what we extract. How often to re-read it depends which field you care about. Size availability turns over fastest, since a single size selling out flips one boolean and nothing else on the page changes. Price and discountRate move on campaign boundaries and during the Spanish sale seasons, which the country config marks with an isDiscountSeason flag. Assortment rotates on a slower cycle as collections drop and older references stop being sellable. Daily suits pricing and stock work, weekly is usually enough for catalogue and assortment tracking, and we set the cadence per field group rather than crawling everything at the fastest rate.