Working with matches
Read how Kabaido rates a match against your catalogue and adjust it when you disagree.
Once your catalogue is loaded, Kabaido matches each line against your products and shows how well each candidate fits, scored from 0 to 100.
Fit ratings
The fit number is calculated deterministically from how well the product satisfies the line, field by field. The same line and the same catalogue always produce the same score.
Each schema field carries a weight, and the fit is the weighted average of how well the product satisfies each requested field, scaled to 100. An exact match scores full marks, within tolerance scores 0.85, an accepted alternative 0.6, a value the product does not state 0.35 and a miss 0. A miss on a hard role field disqualifies the product entirely.
The per-attribute breakdown
Below the fit number, breakdown chips show why the product scored as it did, one chip per attribute:
- Green: an exact match on that attribute.
- Blue: within tolerance.
- Amber: an acceptable alternative, such as an equivalent coating.
- Red: a miss on that attribute.
- Grey: not stated, so unknown.
The review band
A clear winner is selected for you and shown as a high match. When the top candidates are close or none scores highly enough, Kabaido presents the top three for you to review and pick from rather than choosing silently. The bands are fixed:
| Band | Condition | What you see |
|---|---|---|
| Auto select | Fit 86 or higher and at least 8 points clear of the runner-up | One selected product with the alternatives a click away |
| Review | Fit 60 or higher but no clear winner | The top three candidates for you to pick from |
| No match | Fit below 60 | The line stays structured and unpriced, ready for a manual decision |
Tolerances and equivalents
Numeric attributes match within the tolerance set on the schema field, defaulting to 5 percent where the request states none. Fields marked nearest above accept the next size up within a 20 percent overshoot. Known equivalent values, such as TiAlN and AlTiN coatings, count as acceptable alternatives and always show as amber chips, never silently as exact.
Where the AI does and does not touch matching
Scoring and band decisions are pure calculation. The AI reranks candidates only inside the review band, where humans pick anyway, and an exact SKU mention locks straight to that product. A match can always be explained chip by chip.
Challenge or adjust a match
You can select a different candidate from the review list at any time. Use Challenge this match when you disagree with the selection and Kabaido takes a second, deeper look at the candidates with the line context.