Data quality and completeness

Ask three people in the same business whether a product record is complete and you will get three answers. That is not a communication failure — it is a genuine absence of a definition, and until somebody makes the decision explicit the question cannot be answered.

Completeness is a decision, not a fact. There is no universal standard, because a complete record for a t-shirt and a complete record for a variable-speed drive have almost nothing in common. Any vendor offering you a universal completeness standard is offering you somebody else's decision, and it will be wrong for at least half your catalogue.

The workable unit is the category, because within a category buyers ask the same questions. Work through your top categories by revenue and, for each, answer one question: what does a buyer need to know before they will commit? Ask the people who take the calls — counter staff know exactly which question comes up every time, because they answer it every time. That question is a required field.

Three properties separate a rule that is followed from one quietly ignored. It must be checkable without judgement — "has a datasheet" is checkable, "has a good description" is not. It must be visible where the work happens, not in a document nobody opens. And it must be achievable: if a rule cannot be satisfied for a whole brand because the manufacturer does not publish that attribute, the rule is wrong, not the data.