Connecting a supplier normally starts with a feed: XML, YML, CSV or Excel. But a supplier is not obliged to make one, and quite often there is only a website with a catalogue. That is solvable — differently.
Why a feed is simpler
A feed is already structured: SKU, title, price and stock each have their place, and the program only has to read them. The reading rules are the same for every feed of that format.
A website does not have that. It has pages where the price is mixed into the markup, where one supplier puts attributes in a table and another writes them into the description, and where stock is the word “available”. The parsing has to be written for that specific site, and it will not work on another.
What “collecting from the site” means
The program walks the catalogue page by page and takes out the same things it would take from a feed: title, SKU, price, availability, photos, attributes. From there the product goes down the same pipeline as any other — markup, replacements, images, meta tags. At that point there is no difference left.
What the time and price depend on
- how many kinds of pages — catalogue, product card, brand page: each is parsed separately;
- how you walk the catalogue — ordinary pagination or loading as you scroll;
- where the attributes live — a table can be parsed, prose in the description mostly cannot;
- whether the catalogue is closed — behind a wholesale login or a rate limit.
That is why we do not name the price in advance: first we look at the catalogue, then we count. The order of magnitude is on the Pricing page.
What breaks most often
Supplier sites get redesigned without warning, and the parsing stops matching the markup. That is not a fault of the program but a property of the method: any collection from pages lives exactly until the next redesign. So if the supplier has any file at all — even a crooked Excel sent by e-mail — it will be both cheaper and steadier.
How to find out whether it will work
Send a link to the catalogue. We look at how it is built and answer within a day: whether it works, how long it takes and what comes out. Before any payment.