Normalize recurring supplier data without rebuilding the process every time.
Profile known sources, map supplier fields into a stable canonical model, apply reusable transformations and preserve the evidence behind every selected value.
manufacturer_name→product.brandaliasmfr_part_no→variant.mpnnormalizeproduct_group→product.categorydictionaryweight_grams→attribute.weightunit ruleThe file format is only the surface problem.
Recurring supplier data breaks when structure, terminology and meaning drift independently. The platform keeps those concerns explicit instead of burying them in one import script.
Schema profiling
Fingerprint source structure and inspect types/samples before mapping.
Versioned mapping
Keep field mappings and deterministic transformation rules as reusable, versioned configuration.
Normalization dictionaries
Turn recurring brand/category aliases into governed decisions instead of repeated manual cleanup.
Source lineage
Keep the original supplier value and import provenance connected to canonical catalog state.
The next file should be a comparison, not a new project.
When the same supplier returns, the application can reuse known mappings and measure whether the source is still compatible with the expected contract.
- Detect additive or material schema changes.
- Reuse the current mapping profile when the schema still fits.
- Quarantine critical delivery drift when configured reliability controls require it.
- Preserve every import as separate operational history.