Product data quality
Product data quality is the degree to which product records are complete, accurate, consistent, valid, timely and unique enough to support the decisions and systems that depend on them. In commerce it is measured against a defined attribute standard per category rather than in the abstract — a record can be flawless as data and still be unfit for purpose if it lacks the specifications buyers filter and choose on.
The six dimensions, applied to a catalog
Data quality frameworks generally name the same dimensions. What varies is what they mean for products:
| Dimension | In a product catalog |
|---|---|
| Completeness | The required attributes for the category are populated |
| Accuracy | Values match the manufacturer's source document |
| Consistency | The same fact is expressed the same way across suppliers and SKUs |
| Validity | Values conform to type, unit and allowed-value rules |
| Timeliness | The record reflects the current revision, price and availability |
| Uniqueness | One SKU is one record, with duplicates resolved |
Consistency is the one most often underestimated. Three suppliers submitting 1/2", 0.5 in and 12.7 mm produce a catalog that is complete, accurate and unusable for faceted search, because the facet fragments into three values that no buyer will click through.
Measure against a standard, not in the abstract
"Our data is 78% complete" is not a measurement until someone says complete against what. Fill rate only means something relative to a defined attribute set per category — the fields a buyer in that category needs in order to choose.
Two refinements make the number honest:
Required-attribute fill rate. Measure against the required set, not every field in the model. Filling optional fields inflates completeness without changing a single purchase decision.
Revenue weighting. A catalog at 71% overall can sit at 34% in the three categories driving most of the margin. Unweighted averages systematically hide the gaps that cost the most, because slow-moving long-tail SKUs dominate the count.
Leading and lagging indicators
Quality metrics are only useful if they connect to something the business already tracks.
Leading: required-attribute fill rate by category, sample accuracy against source documents, unit and vocabulary consistency, percentage of SKUs with valid GTIN/MPN, structured-markup validity, age of the newest supplier revision.
Lagging: null-result search rate on site, facet abandonment, marketplace and retailer rejection rate, returns coded as "not as described", quote and RFQ cycle time in B2B, and conversion on enriched versus unenriched SKUs.
The pairing matters. Fill rate on its own reads as a hygiene project; fill rate next to null-result search rate reads as revenue, which is the version that survives a budget review.
Frequently asked questions
What is a good product data quality score?
There is no universal threshold, because the target depends on category and channel. A useful working standard is 100% on the attributes a buyer filters on in that category, and best-effort on the rest. Chasing a high catalog-wide average usually means filling easy optional fields while the deciding attributes stay empty.
How is data quality different from data governance?
Quality is the state of the data; governance is the set of rules, owners and processes that keep it in that state. Governance without measurement produces policy nobody can verify, and measurement without governance produces a number that improves once and then decays as new SKUs arrive.
Who owns product data quality?
In practice it is shared between merchandising, ecommerce and IT, which is why it often ends up owned by nobody. The arrangement that works is a named owner for the attribute standard per category — usually a merchant with category knowledge — and a separate owner for the pipeline that fills and validates against it.
Does a PIM guarantee data quality?
No. A PIM enforces structure, validates rules and makes gaps visible and measurable, which is genuinely valuable. It does not research missing values or verify that a stated specification matches the manufacturer document. Fill rate immediately after a migration is usually close to what it was before it.