Essential metrics and KPIs for measuring data quality
Measuring data quality is not optional when operating high-risk AI systems. AI Act Art. 10 requires relevant, representative and error-free data. Without clear metrics, you cannot demonstrate it.
Why you need data quality KPIs
A Data Quality Management programme without indicators is unmanageable. Data quality KPIs allow you to:
- Establish a baseline and measure evolution over time
- Prioritise which data domains require urgent intervention
- Report to the Management Committee and auditors with objective data
- Demonstrate compliance with AI Act Art. 10 and GDPR Art. 5.1.d
The 6 quality dimensions and their metrics
1. Completeness
What it measures: the percentage of records with all required fields covered.
Formula: (Number of completed fields / Total required fields) × 100
Recommended threshold: ≥ 95% for high-risk AI training data.
2. Accuracy
What it measures: the percentage of records whose value matches reality or an authoritative reference source.
Formula: (Accurate records / Total records) × 100
Recommended threshold: ≥ 98% for critical business attributes.
3. Consistency
What it measures: the percentage of records with the same value for the same concept across different systems.
Formula: (Consistent records across systems / Compared records) × 100
Recommended threshold: ≥ 99% for master data.
4. Uniqueness
What it measures: the percentage of records without uncontrolled duplicates.
Formula: (1 - Duplicate records / Total records) × 100
Recommended threshold: ≥ 99.5% for master data.
5. Timeliness
What it measures: the percentage of data available within the defined SLA timeframe.
Formula: (Data available on time / Total expected data) × 100
Recommended threshold: ≥ 98% for critical operational data.
6. Validity
What it measures: the percentage of records that comply with defined business rules and formats.
Formula: (Valid records / Total records) × 100
Recommended threshold: ≥ 97% for regulated data.
Programme KPIs (management level)
In addition to dimension metrics, you need programme KPIs to report to the CDO and Management Committee:
- Overall Quality Index (OQI) — weighted average of the 6 dimensions
- Number of open quality incidents — by domain and by criticality
- Mean Time To Resolution (MTTR) — of quality incidents
- Quality rule coverage — % of critical attributes with defined rules
- % data domains with assigned Data Owner
How to implement tracking of these KPIs
The recommended process in 4 steps:
- Inventory: identify critical domains and attributes
- Rules: define quality rules for each attribute
- Measurement: implement controls in the data pipeline
- Dashboard: visualise results and generate automatic alerts
What's your Data Governance maturity?
Free assessment with your priority gaps, plus the self-assessment quiz and savings calculator on the Data Governance path.