Sources never meant to be integrated
Wholesalers in this sector do not offer tidy APIs: the information lives in web listings and catalogs written for a person to read.
Five wholesalers wired into one store: integrations that read each supplier and leave finished products behind, an AI pass that files them under the right category, and a server holding the company’s three sites.
Every garment multiplies by its colors and its sizes, and every combination carries its own code and its own stock.
Panda Merchandising sells promotional products and corporate clothing. Its business has one feature that defines everything else: the catalog is not its own. Panda resells wholesaler stock, each supplier holding thousands of products that exist in dozens of color and size combinations, with prices and availability that change constantly. Loading that by hand is not viable: one product can mean a listing, six photos and more than twenty variants with their own code and stock. Five wholesalers feed the shop today, and none of them publishes a tidy API for it.
Wholesalers in this sector do not offer tidy APIs: the information lives in web listings and catalogs written for a person to read.
Dumping data is not enough. The product needs its photos, its variants correctly associated and complete information before it can sell.
The client had to be able to refresh the catalog without a technical intervention each time.
The importers were the real heart of the project: integrations that read a supplier’s catalog and leave a finished product inside the store.
Five suppliers, five integrations: ZECAT, Stocksur (formerly CDO), PromoProductos, Bello Export and Carhue. Each one reads its own source and leaves a finished product in the store, with the listing, the description, the image gallery, the color and size attributes and the full variant matrix, every combination carrying its own code and stock state.
An importer knows what a product is called, not where it belongs in a shop with 240 categories. A plugin of our own sends each unfiled product to Gemini with the category tree and writes back the ones it picks, weekly, in batches, and it never overrides a category a person chose.
Imports are checked against the source to catch missing variants, incomplete codes or images that did not come down. A separate nightly guard goes further: a product that has disappeared from its supplier is unpublished instead of left quietly buyable.
The full site, shop and institutional presentation included, and later a migration that brought the company’s three sites onto one server we run, so the catalog machinery and the hosting answer to the same people.
Only what this project actually runs on.
Panda can widen the offer, keep stock aligned with the wholesaler and respond quickly when a new line appears, without that meaning weeks of administrative work. The catalog stands at 1,610 published products and 3,940 variants across five suppliers, filed into 240 categories. The automatic passes close the loop at both ends: every import is checked against its source, and anything the supplier drops comes off the shop the same night.
Tell us what you are building and where it is stuck. We answer within one business day.