Data Metrics¶
Measure the health of your data products, not just your pipelines.
Data Quality Metrics¶
| Metric | What it measures |
|---|---|
| Freshness | How recent is the data? Is it within SLA? |
| Completeness | What percentage of expected records are present? |
| Null rate | How often are critical fields null? |
| Uniqueness violations | Duplicate records on key fields |
| Business rule failures | Records failing domain validation |
Pipeline Reliability Metrics¶
- Pipeline success rate — percentage of runs completing without failure
- Mean time to detect — how quickly are data issues spotted?
- Mean time to resolve — how quickly are data issues fixed?
- SLA attainment — is data available when consumers expect it?
Data Volume Metrics¶
- Row count trends — are volumes within expected ranges?
- Anomaly detection — sudden drops or spikes
- Schema change frequency — how often does structure change?
Suggested Tooling¶
| Tool | Purpose |
|---|---|
| dbt tests | Schema and business rule validation |
| Elementary / re_data | dbt-native data observability |
| Monte Carlo | Data observability platform |
| Great Expectations | Expectation-based data validation |
| Soda | SQL-based data quality checks |
Use what fits the stack and the team. The important thing is that quality is measured automatically, not manually.