A practical checklist for data quality assessments
Validity, reliability, timeliness, precision and integrity — what to check in each, and what to do when something fails.
Data quality assessments are often treated as compliance exercises. Done well, they are one of the most useful things a programme can do: they show where numbers can be trusted, and where reported results may be telling a misleading story.
Validity
- Does the indicator definition match what is actually counted in the field?
- Are the data collection tools asking the right question, in the right way?
- Is anyone counted who shouldn't be — or missed who should be?
Reliability
- Would two people collecting the same data reach the same number?
- Have tools, definitions or methods changed between reporting periods?
- Are source documents filed and retrievable?
Timeliness, precision and integrity
- Is the data available in time to inform the decisions it is meant for?
- Is the margin of error small enough to detect the change you expect?
- Are there safeguards against transcription errors and manipulation?
When something fails
Every weakness found should come with a proportionate fix and an owner. A short improvement plan, revisited at the next assessment, is worth more than a long list of findings.