Definition
Product attributes go missing for four of the most common structural reasons: the supplier never sent the data, the schema requires a field no upstream system actually holds, the attribute was added after the products were onboarded, or nobody owns the backfill. A fifth cause sits in transit rather than at the source — the supplier did populate the field, but mapping or transformation between systems dropped it, so the value exists upstream and the catalog is blank. Each of these is a different failure with a different fix, and several are usually present at once in the same catalog. Blank fields are the symptom they all share, which is why counting blanks never tells you which one you have.
Key points
- Supplier incompleteness: the data never existed in the file you received, so no amount of downstream processing can recover it.
- Schema-source mismatch: a field is marked required but no system of record captures it, so it is unfillable by design.
- Retroactive schema change: the attribute was added later and existing products were never revisited, so gaps cluster in older records.
- No backfill ownership: the gap is known, fillable, and unassigned, so it survives every quarter it is reported in.
- Mapping loss in transit: the value arrived but attribute mapping between systems dropped or misrouted it, so it is present in the source file and absent from the catalog record.
How do you tell which cause you actually have?
Start with the cheapest test: diff a sample of catalog records against the supplier files they came from. A value present in the file and blank in the record is mapping loss, not missing data, and no enrichment budget should be spent on it. Then segment the remaining blanks rather than totaling them. An attribute empty for two suppliers and full for the rest is a supplier-file problem, and the fix is an onboarding requirement. Gaps confined to products onboarded before the attribute was defined point at a retroactive schema change. An attribute empty across every supplier and every vintage means no source holds it. Anything that survives all four cuts and keeps reappearing in reports is an ownership gap.
Common pitfalls
- Fixing the symptom in the outbound feed rather than the catalog, so every channel needs its own patch while the canonical record stays empty.
- Filling gaps with filler values such as "N/A" or "Standard", which clears the completeness report while carrying no information at all.
- Treating a rising completeness score as the goal, when the score can rise without a single answerable question becoming answerable.
- Running an enrichment project before segmenting the causes, so effort goes to gaps that were never fillable from any available source.
FAQ
Isn't this really just the attribute schema being wrong?
Sometimes — a schema requiring fields no source populates is one of the causes. But the others sit outside it: upstream at the supplier, in the pipeline between systems, or in a process that never assigned the backfill. Rewriting the schema will not recover data a supplier never sent, will not restore a value the mapping dropped, and will not give a backfill an owner.
Can attribute extraction fill the gaps automatically?
It can recover characteristics that are stated somewhere unstructured — in a description, a spec sheet, or an image. It cannot produce a characteristic that was never captured anywhere. That boundary is the practical dividing line between the supplier-incompleteness cause and the schema-source-mismatch cause.
Source
A worked example of category-level required attributes — the standard whose absence produces the schema-source mismatch described above — is Google's Merchant Center product data specification, which states for each field whether it is required, optional, or required only under specific conditions.