Data quality is often discussed as a property of a dataset. In a working system, it is also a property of the steps that connect the dataset to a decision. Downloading, cleaning, matching, aggregating, modelling, and presenting information each introduces choices that need attention.
Preserve the source context
Keep the provider, version, publication date, licence, variable definition, units, and relevant limitations with the data. A file name rarely captures enough context. Where a product is updated, record which version supported an earlier result. This makes it possible to investigate differences rather than guessing why two outputs disagree.
Inspect the transformations
Unit conversions, coordinate transformations, joins, and missing-value rules can all create silent errors. Test these steps with examples that have known outcomes. Check edge cases around boundaries and incomplete records. Document why an aggregation method was chosen and what detail is lost when individual observations become a portfolio-level summary.
Communicate remaining limitations
A reliable process does not make uncertainty disappear. It makes the uncertainty traceable and its consequences understandable. Show coverage gaps and confidence limits where they matter. Define review triggers for source changes or unexpected patterns, and ensure that technical findings can reach the people responsible for the decision.
Three things to take away.
- Track provenance through every transformation.
- Test joins, conversions, and missing-data rules.
- Make remaining uncertainty visible to decision owners.
Further reading
Explore the underlying topics through these reference sources.