Keep a catalog in sync automatically

Before touching ProcuLink, establish three things about the supplier's file, because they decide every setting that follows:

• How you reach it — an HTTPS URL, or a host and path on SFTP/FTPS/FTP.
• What authenticates — nothing at all, an API key header, a bearer token, basic auth, or OAuth2 client credentials.
• What shape it is — CSV, XLSX, JSON, XML, or an Ariba CIF file. And whether it arrives inside a ZIP, which is common for large distributor price lists.

You do not have to be right about the shape. ProcuLink's Auto-detect handles the common cases including ZIP extraction, and step 4 shows you what it actually found. But you do have to be right about the address and the credential.

Open Partners → Suppliers → the supplier → Catalog, then expand Keep the catalog in sync automatically — "Pull on a schedule or let the supplier push — instead of uploading by hand."

The panel is headed Pull from a file server or API: "ProcuLink fetches the supplier's catalog on a schedule and upserts products by code."

Choose the protocol first, because it changes the fields below it:

Then set File format. Leave it on Auto-detect unless a previous attempt got it wrong — the explicit choices are CSV, XLSX, JSON, XML (cXML Index or vendor XML), and CIF (Ariba 3.0).

Under Authentication, pick the method the supplier told you about and fill in its fields. Secrets are write-only: once saved, the field shows a masked placeholder and leaving it blank keeps the stored value.

Press Save source.

This ordering is not optional and it catches everybody: Test connection & preview stays disabled until a source has been saved, and the button says so on hover — "Save the source first, then you can test the connection." The test runs server-side against stored credentials, so there has to be something stored.

Leave Enabled switched off for now. You do not want a schedule running against a configuration you have not proven.

Press Test connection & preview. This is read-only — it connects, downloads, and parses, and it writes nothing. The panel says as much: "This is a read-only preview — it connects, reads the file, and shows what would be imported. It does not save anything."

A success line reads Connected. followed by the file name, the detected format, the byte count, the row count, and how many of those rows have a code. Below it:

• Mapped columns — their column name, and the ProcuLink field it became.
• Ignored columns — everything the parser saw and is not reading. "ProcuLink reads only the mapped columns above."
• Sample rows — the first few rows as they would be imported.

Now check three numbers, in this order:

1. Rows with a code. If this is zero, nothing would import. Go to step 5.
2. Is price in Mapped columns? A catalog with codes and names but no prices is a common and quiet failure — the supplier's price column is often named something the aliases do not cover.
3. Do the sample rows look like products? Read five. A price that is 100× out, or a name that is obviously a category, means a column landed in the wrong field.

Most feeds need this once. Open Map columns (advanced) — "Only needed when the supplier's column names aren't recognised."

If the last preview found columns it could not place, they appear as amber chips under From the last preview — click to map. Clicking one adds a mapping row for it, which is much faster than typing names that are usually in another language.

Each row is Their column (or index) → ProcuLink field. Then Save source and test again.

Two switches at the bottom of that section cover the awkward feeds:

• This file has no header row (map by column position, starting at 0). Some distributor exports are raw data with no header at all. Turn this on and map by index: 0 → code, 3 → price, and so on. The sample rows from the preview tell you which index is which.
• File encoding. Leave blank for UTF-8. A legacy export with mangled accented names usually wants windows-1252.

Feed shapes worth recognising, because each needs a different one of the above:

Once the preview is right, switch on Enabled, set Every (hours), and Save source.

The helper text describes the behaviour exactly: "ProcuLink checks this often and only re-imports when the file has changed." So a daily catalog on a 6-hour schedule costs three cheap checks and one real import, not four imports. Anything from 1 hour to 336 (two weeks) is allowed; for a price list that moves monthly, daily is generous.

Two limits to know rather than discover: a feed can be up to 256 MB and 200,000 rows. Every ordinary distributor price list is comfortably inside both.

Come back after the first scheduled run. The Catalog tab shows the last sync's outcome, and the numbers are worth reading rather than glancing at.

A healthy first sync on a fresh supplier reports almost everything created and nothing skipped. A healthy second sync reports almost everything updated — because matching is an upsert on the product code, scoped to this supplier, so re-running does not duplicate anything.

That arithmetic is the check: if a supplier had 38 products imported by hand and the feed carries 14,713, a correct sync reports 14,675 created plus 38 updated, and the catalog total is 14,713 — not 14,751. A total that grew by more than the number of genuinely new products means codes are not matching, and the usual cause is a stray prefix or trailing space in the mapped code column.

Then spot-check the catalog search for five products you know, exactly as you would after a manual import.