Matching resolves a row of imported data to a specific product. It runs before any content is written, because everything downstream inherits the result.
How Matching Works
Matching uses identifiers first, because manufacturer together with model or part number is the strongest signal available, and falls back to broader evidence when identifiers are incomplete or unclear. Each candidate carries a confidence indication and the reason it was proposed.
| Outcome | What it means | What happens |
|---|---|---|
| Confident match | Identifiers resolve to exactly one product. | Proceeds automatically to the next stage. |
| Ambiguous | Several products fit comparably well. | Held for human confirmation with the candidates shown side by side. |
| No match | Nothing fits the identifiers provided. | Sent to exceptions. This is usually a data problem rather than a matching problem. |
Ambiguity Is Information, Not Failure
A row matching three products comparably usually means the identifier is not specific enough to distinguish variants. That is worth fixing in the source data rather than guessing past.
Confirming in Groups
Ambiguity clusters. A supplier file with a systematic formatting quirk produces many ambiguous rows of the same shape, and resolving the pattern usually clears all of them at once. Sorting the ambiguous queue by cause turns what looks like hundreds of decisions into a handful.