Working notes
What makes a product import safe to repeat?
Getting the first import through is only half the job. The next file needs to update the right records without quietly creating a second catalogue.
The short answer
A repeatable import needs stable matching, clear update rules and a second-run test. A CSV importing once is not enough.
Why it matters
A file importing successfully does not prove the process is ready to run again. The next supplier sheet might use a different title, a new column order or an identifier that has changed. I want to know how the next run behaves before calling something repeatable.
Choose an identifier that represents the record
A title is useful for people, but it is often a poor long-term match key. Where reliable identifiers exist, I'd use them and check their uniqueness.
An internal record ID and a reliable product identifier may serve different purposes. Keep the distinction visible instead of assuming one field can stand for a product, a variation and a supplier row at the same time.
Know how the importer matches
An importer's matching rule belongs to that import job. Starting a completely new import does not guarantee it will recognise the same existing records automatically.
For a new job updating existing items, use explicit existing-record matching and test it. For this website, the permanent dm_record_id field is intended to make that matching possible without relying on the page title.
Agree which fields may change
A supplier price update should not silently replace hand-written descriptions or delete images because those columns are missing. The update rules need to say what the file owns.
Blank, missing and intentionally cleared values should also be treated deliberately. They are not always the same instruction.
Tell me what you need to sort out
If you need this checked or fixed on a real site, send me the URL, a safe example and what you are trying to achieve.
Run the same sample twice
Check a small representative batch, then repeat it. Look for duplicates, lost values and records that should have matched but did not. Keep an exception list rather than guessing where a match is ambiguous.
Price-import checks and the difference between an import and a sync cover the next parts of the process.
HAVE SOMETHING AWKWARD?
Tell me what you need to sort out
Send me the URL, what is happening, what you have already tried and what a useful result would look like. A small anonymised sample is enough to start - do not send passwords, credentials or sensitive customer data.