Home › Data quality
What is data quality, and how do you measure it?
Data quality is the extent to which your data is usable for the decision you make with it: is it correct, is it complete, and does it mean the same thing everywhere. You measure it with a baseline per source and per definition, and you keep it up by recording who owns which data rather than by buying a tool.
Start with a baseline, not a tool
Pick the figures people actually steer on. For each one, establish where it comes from, how complete it is, and who decides what it means. That produces a list you can act on, instead of a score nobody knows what to do with.
The four questions per figure
- Where does it come from, and through how many hands?
- How often is it missing or obviously wrong?
- Who decides what it means, by name?
- What happens if it is wrong, and who notices?
The last one is the most useful. A figure nobody notices being wrong is a figure nobody is steering on.
Sonny and I collaborated on a data management project to measure and steer data quality through BI dashboards. He delivered a high-quality Power BI template and added value with strong visual storytelling and a custom function for column-level checks.
Half an hour is enough to know whether we fit
No slide deck and no quote at the end. We walk through your sources, your definitions and your biggest frustration.
Book half an hourFrequently asked questions
What is data quality and how do you measure it?
Data quality is the extent to which your data is usable for the decision you make with it: is it correct, complete, and does it mean the same thing everywhere. You measure it with a baseline per source and per definition, and you keep it up by recording who owns which data.
Where do you start?
Not with a tool but with a baseline. Pick the figures people actually steer on, and check for each of them where it comes from, how complete it is and who decides what it means. That gives you a list you can act on rather than a score nobody can use.
Who should own data quality?
Someone in the business, not in IT. The person who suffers when the number is wrong is the one who should decide what right means. IT can enforce the rule, but only after someone has decided what the rule is.
How do you keep it from slipping?
By making the checks part of the daily load rather than a periodic project. A rule that runs every night and reports a deviation the next morning gets fixed. A quarterly report gets filed.
Is this necessary before using AI?
Yes, and it is the least popular part of the answer. A model on messy data gives you wrong answers faster. Without agreed definitions there is nothing for a model to be consistent with.