Data Quality in a Product Passport

Internal data quality problems are absorbed by the people who know the data. A passport removes that buffer, exposing whatever is in the record to consumers, regulators, recyclers and competitors simultaneously.

That changes which quality problems matter and how much effort each deserves.

Missing data is safer than wrong data

A gap is visible, and everybody handling the record knows it is a gap. A wrong value is invisible and is acted upon.

This inverts the instinct on most data programmes, which treat completeness as the primary measure. For a passport, an incomplete record honestly labelled is a considerably better outcome than a complete record containing values nobody has verified.

The four failure modes that matter

Most passport data problems fall into four categories, and they need different detection approaches because only one of them is caught by ordinary validation.

Only the last row is caught by validation, which is where most quality effort goes. The four above it all produce values that pass every schema check and are wrong.

Stale data is the commonest problem

Product data decays. A supplier substitutes a material, a specification is revised, a test is repeated with a different result, a calculation method changes.

A passport built on copies rather than references accumulates these silently, and the record looks exactly as complete and confident when stale as when current.

The practical defence is to record when each value was established and what its expected refresh interval is, then monitor age against that interval. A footprint figure four years old is not necessarily wrong and it should be flagged for review rather than published with the same confidence as one from last quarter.

Outlier analysis catches what validation misses

Unit confusion and misplaced decimals produce values that are structurally valid and physically implausible, which makes them detectable statistically rather than syntactically.

The last step matters. Automatically correcting outliers destroys genuine variation, and some of the most useful findings from this exercise are products that really are unusual for reasons nobody had noticed.

Provenance is the honest answer to gaps

No programme achieves fully verified, supplier-specific, current data across a whole portfolio. The question is what to do about the difference between that ideal and reality.

• Record the source of every value — measured, supplier-declared, calculated, or estimated from a category average.
• Record when it was established, and by whom.
• Record the method where one applies, since a figure without its method is not interpretable.
• Record the evidence reference, so a challenge can be answered.

A value labelled as a category-average estimate is useful and honest. The same value presented as a supplier-specific measurement is a misrepresentation, and the difference between them is entirely in the provenance record rather than in the number.

What to measure instead of completeness

Since completeness targets are counterproductive, a programme needs measures that reward the right behaviour.

Proportion of attributes with recorded provenance is a better primary measure, because it can only be improved honestly. Proportion resting on primary rather than estimated data shows genuine progress. Median age of values shows whether the record is being maintained or merely populated once.

These three together describe a passport programme’s actual health far better than a completeness percentage, and none of them can be improved by filling a field with a guess.

Where to put the effort

Quality effort should follow consequence rather than spreading evenly across attributes, because the consequences of being wrong differ enormously between them.

A wrong substance concentration has safety and legal consequences. A wrong carbon figure is a green claims exposure. A wrong product weight is embarrassing. Ranking attributes by what happens if the value is wrong produces a very different priority list from ranking by how hard they are to obtain, and it is the more useful ordering.