Data can be accurate and complete and poor quality

In the right context, completeness and accuracy are key data quality indicators.

Data quality has various definitions. Many of them focus on completeness and accuracy. So DQ metrics often focus just on whether data is complete and accurate.

But if we use data that is not relevant to our context, these metrics don’t matter. The data might be 100% accurate and 100% complete, but still meaningless. This means that, in the context of our use case, the data isn’t actually high quality for what we need.

When deciding how to measure data quality, the first question should really be “Is this data fit for our purpose?”

Examples in practice

1. External demographic data from third parties like Experian. We explored this in this article about causation and this article about gender data. These might be useful for certain marketing activities, but we should not be using them to price risk, approve loans or adjudicate claims. Even if a data provider reports “high quality”, it’s not relevant to our context and is therefore poor quality for our purpose.
   
   
2. Internal data that is not relevant to our use case. We explored driver’s license type in this article about the direct monetary impact of bias. In cases like these, the data might be accurate, but not relevant. In fact, making it less “accurate” could actually make it higher quality for our purpose.

Dedicated data quality or governance teams may not spot the difference. Context matters, and this often means that input from subject matter experts is needed. You can’t completely outsource your DQ responsibilities to a central team.

Disclaimer: The info in this article is not legal advice. It may not be relevant to your circumstances. It was written for specific contexts within banks and insurers, may not apply to other contexts, and may not be relevant to other types of organisations.