Data Validation

What is Data Validation used for?

Data Validation checks the quality and usability of active Smart Products. It identifies missing, invalid or improvable values. The workflow is shared by all sources.

A validation finding does not mean that import failed. A technically imported product may still need content attention.

When are checks run?

Validation results are recorded while products are processed. Rescan recalculates checks for already stored active Smart Products; it does not fetch the source, start an import, write to a store or automatically apply fixes. Import first when the source itself changed.

Score and counts

The summary shows checked products, error-free products, critical findings and a directional quality score. This is not channel approval: sales channels and categories may impose stricter requirements.

Product and issue counts differ. One product can have several findings and one grouped issue can affect many products.

• Error: missing or invalid information with strong impact, such as invalid price or GTIN.
• Warning: requires attention but does not automatically make import unsuccessful, such as missing or duplicate GTIN.
• Information: improvement or explanatory notice.

Common checks cover titles, descriptions, GTIN, price, availability, stock, brand, product type, images and links.

Find and resolve issues

Search and filter by severity or resolver. Each grouped row shows affected products, finding, impact, solution type and action.

• Manual: factual data that Nifelo must not invent, such as identifiers, prices, stock or wrong mappings. Fix structural errors in the authoritative source.
• Knowledge: deterministic correction using managed reusable knowledge.
• AI: suitable content issues. AI is not appropriate for unique identifiers or other hard facts. Always review results.

AI jobs

Single or bulk AI repair may be available for eligible findings. AI settings, monthly and release limits, wallet permission and credits apply. Conflicting work can be blocked during import and larger jobs run in the background. Do not restart a job already queued or processing. Successful AI changes receive protected field ownership.

The page shows queued, running, completed or failed jobs and changed, unchanged or failed products. Completion means processing ended, not that human review is unnecessary.

Image scan and price mapping

The separate image scan checks image references in the background and is distinct from AI Vision. Fix broken URLs in the source or mapping.

For a price issue caused by mapping, Nifelo may offer source-field candidates. Verify the real selling-price field. After changing a mapping, a new import is normally required.

Ownership and workflow

Customer edits and successful AI values can remain protected field by field during later Smart Update imports. Use the source for factual errors, Smart Fields for mappings, rules for predictable exceptions, knowledge for managed corrections and AI for appropriate content.

Recommended order: verify technical import, solve errors first, correct source and mapping problems, use the appropriate resolver, monitor background tasks, import after source changes, rescan stored products and finally check Coverage & analysis and the target channel.