The 6 Dimensions of Data Quality — Accuracy: The Data Is Clean. It's Also Wrong.
Data accuracy needs a real-world reference. Reconcile metrics with payments, receipts, and domain knowledge before trusting a tidy dashboard.
Read the essay← All categories
Category
Data Quality
Recognize the failures that make tidy, persuasive data lead to the wrong decision.
Start with completeness: missing data does not always look broken.
Reading guide
What questions are we asking?
Data quality starts with “good enough for which decision?” A table can contain valid-looking values and still omit a whole device, period, or customer group. Start with completeness and learn to distinguish a missing observation from an observed zero.
- Is the data complete for this question, at this grain and over this period?
- Are missing records being interpreted as zero activity?
- Could a tracking change explain the apparent business change?
A suggested reading path
The 6 Dimensions of Data Quality — Completeness
Start with coverage gaps, MCAR/MAR/MNAR, and the business consequences of missing data.
The 6 Dimensions of Data Quality — Validity: A Valid Email Address Can Still Be Terrible Data.
Define the format and business rules that a populated field still needs to satisfy.
The 6 Dimensions of Data Quality — Uniqueness: One Customer, Four Rows, Zero Trust.
Choose the counting unit before treating repeated rows as duplicates.
The 6 Dimensions of Data Quality — Consistency: When Two Dashboards Disagree, Neither Team Is Crazy.
Align definitions, filters, and ownership before comparing dashboard totals.
The 6 Dimensions of Data Quality — Timeliness: Yesterday's Data Is Not Yesterday's Truth.
Distinguish a current number from a complete, decision-ready number.
The 6 Dimensions of Data Quality — Accuracy: The Data Is Clean. It's Also Wrong.
Reconcile the result with the real-world process it claims to measure.
Try it at work
Before filling nulls, make a coverage table by date and by the segment your decision depends on. Compare it with an independent source and investigate missing slices. Then use the other five dimensions to check rules, counting units, definitions, timing, and agreement with reality.
Series
Data Quality Is Six Different Ways to Make a Bad Decision
Six dimensions of data quality in business terms: completeness, validity, uniqueness, consistency, timeliness, and accuracy.
Explore the series →All essays in this category
The 6 Dimensions of Data Quality — Consistency: When Two Dashboards Disagree, Neither Team Is Crazy.
Conflicting dashboards often hide different metric definitions. Align grain, filters, timezones, and ownership with a practical metric contract.
Read the essayThe 6 Dimensions of Data Quality — Timeliness: Yesterday's Data Is Not Yesterday's Truth.
Fresh data can still be incomplete. Define latency, reporting windows, and freshness expectations before treating provisional metrics as final.
Read the essayThe 6 Dimensions of Data Quality — Uniqueness: One Customer, Four Rows, Zero Trust.
Duplicate events can manufacture growth. Define the counting unit, audit candidate keys, and investigate the mechanism before removing repeated rows.
Read the essayThe 6 Dimensions of Data Quality — Validity: A Valid Email Address Can Still Be Terrible Data.
Data validity goes beyond accepted formats. Define business rules, controlled values, and responses to violations before trusting a field for analysis.
Read the essay
The 6 Dimensions of Data Quality — Completeness
Missing data can look perfectly normal. How completeness failures distort analysis and the decisions that follow.
Read the essay