Data Quality (KA 11) — The Six Dimensions

Data Quality (KA 11) — The Six Dimensions

Accuracy, completeness, consistency, timeliness, uniqueness, validity — measure, target, monitor, escalate.

Why it matters

DQ is the KA most often equated with ‘a dashboard’ and most rarely run as a programme. DMBOK organises it around measure → target → monitor → escalate, the same loop SRE uses for service reliability.

Going deeper

Six DQ dimensions, each with a typical SQL-able rule:

Dimension	Definition	Sample rule
Accuracy	Does the value reflect reality?	Cross-check against system of record
Completeness	Are required fields present?	COUNT(*) WHERE col IS NULL = 0
Consistency	Same fact agrees across systems?	SUM(a.value) = SUM(b.value)
Timeliness	Fresh enough for use?	max(updated_at) > now() - interval '1 hour'
Uniqueness	One row per real-world entity?	COUNT(*) = COUNT(DISTINCT key)
Validity	Does it satisfy declared format/range?	regex / CHECK constraints

Pick a target per dimension per dataset (perfection is uneconomic), instrument, and route the alerts to the right human. Without that last step the dashboard is decorative.

Data quality is the six vital signs on a hospital chart.

Pulse, blood pressure, respiration, temperature, oxygen, level of consciousness — any one out of band escalates differently. Nobody averages vital signs into a single score, because the response depends on which sign is failing.

DQ is the same. ‘Overall data quality is 87 %’ is meaningless; ‘completeness on customer_email is 60 %, uniqueness on order_id is 99.9 %’ tells you what to fix. Always score and alert per dimension.