Data Collection · Data quality · Component
The measures
Completeness, accuracy, timeliness and consistency, per data set — measured rather than assumed.
The deliverable
What it is
Quality is assumed adequate until something visible goes wrong. Measuring it in advance costs little and prevents the class of error where a decision rests on a field populated in half the records.
Completeness is the cheapest to measure and the most often failing, which makes it the place to start.
One level in
What it is made of
Each element is a constituent part of the component. Follow one to see the attributes it carries.
The completeness
What proportion of records have each field.
3 attributes: Populated · Field · Usable threshold
LearnThe accuracy
Whether the values are right.
3 attributes: Accuracy rate · Sample size · Verified against
Learn
Internal consistency is not accuracy. Sample against something external, however small the sample.
The other components in data quality
The causes
Where quality problems originate, which is usually one step in collection rather than many small failures.
LearnThe fixes
What was changed at source, rather than corrected downstream — because a cleaning step fixes one report and source fixes all of them.
Learn