Marketing Operations
RGM° · Training
Data Quality and Governance
Unglamorous and the most leveraged investment in any analytics program. Six dimensions, lifecycle, testing, monitoring, MDM.
Why data quality matters
Bad data flows downstream silently. Reports show wrong numbers; algorithms learn from noise; decisions optimize the wrong outcomes. Data quality is unglamorous and the most leveraged investment in any analytics program.
Six dimensions
| Dimension | What it means |
|---|
| Accuracy | Values reflect reality |
| Completeness | No missing required fields |
| Consistency | Same value across systems |
| Timeliness | Data is fresh enough |
| Validity | Conforms to expected format/range |
| Uniqueness | No unintended duplicates |
Lifecycle
- Collection. Validate at point of capture.
- Ingestion. Test schemas at warehouse load.
- Transformation. Tests on derived tables (dbt tests).
- Consumption. Anomaly detection in dashboards.
- Feedback. Issues reported and tracked.
- Remediation. Root cause fixed; not just patched.
Automated testing
- Schema tests. Column types, nullable, allowed values.
- Uniqueness tests. Primary keys actually unique.
- Referential integrity. Foreign keys resolve.
- Freshness tests. Data updated within expected window.
- Volume tests. Row counts within expected range.
- Business logic tests. Domain-specific validity (e.g., revenue > 0).
- Tools: dbt tests, Great Expectations, Monte Carlo.
Production monitoring
- Real-time alerts on data quality breaches.
- Daily summary of test passes/failures.
- Anomaly detection on key metrics.
- Data SLAs with owners and response times.
- Incident response runbook.
- Postmortems for significant issues.
Cataloging and lineage
- Data catalog. Searchable inventory of tables, columns, definitions.
- Lineage tracking. Where data comes from; what depends on it.
- Ownership documented. Who maintains each dataset.
- Tools: dbt docs, Atlan, Alation, Collibra.
- Self-service. Stakeholders find their data without asking.
Privacy and compliance
- PII handling documented.
- Deletion requests propagate across systems.
- Retention policies enforced.
- Access controls role-based.
- Consent management integrated.
- GDPR, CCPA, sector-specific compliance.
- Audit trails for sensitive data access.
Master data management
- Customer master. Canonical customer identity across systems.
- Product master. Single source of truth for SKU data.
- Reference data. Standardized values (country codes, currencies).
- Hierarchies. Organizational structures, product categories.
- Stewardship. Roles responsible for master data accuracy.
Advanced playbook
- Data contracts. Formal agreements between producers and consumers; enforced in CI.
- Shift-left testing. Tests at producer (source application) not consumer (warehouse).
- Anomaly detection ML. Beyond rule-based tests; ML-detected pattern breaks.
- Data observability platforms. Monte Carlo, Bigeye, Soda — comprehensive monitoring.
- SLA tiers. Critical data has hour-level SLA; secondary has day-level.
- Quarterly DQ review. Top issues, root causes, prevention plans.
- Cross-functional ownership. Engineering owns producers; analytics owns warehouse; together own quality.
- Documentation as deliverable. Data catalog kept current alongside code.
- Incident postmortems. Each significant DQ incident reviewed for prevention.
- Annual privacy compliance audit. External or internal review of data handling.
Common mistakes
- No automated testing; issues found late.
- Tests at consumer only; producer issues missed.
- No data ownership; orphaned datasets.
- Incident remediation patches symptoms; root cause unaddressed.
- Catalog missing; nobody knows where data is.
- Lineage not tracked; impact of changes unknown.
- Privacy compliance reactive; audit reveals gaps.
- Master data inconsistent across systems.
- No incident postmortems; same issues recur.
- Access controls coarse; sensitive data over-shared.
- Retention policies undocumented.
- Data quality metrics not reported to leadership.
Operating checklist
- Automated data quality testing (dbt tests, Great Expectations, etc.)
- Production monitoring with alerts
- Data catalog current
- Lineage tracking
- Data ownership documented
- SLA tiers documented
- Privacy compliance audit annually
- Master data management discipline
- Incident postmortems for significant issues
- Quarterly DQ review
- Data contracts where applicable
- Access controls role-based
Sources and further reading
- Monte Carlo — data observability methodology
- Bigeye, Soda — data quality platforms
- dbt tests documentation
- Great Expectations documentation
- Atlan, Alation, Collibra — data catalogs
- Chad Sanderson — data contracts
- David Jayatillake — data quality writing
- DAMA International data governance
- Modern Data Stack community
- GDPR and CCPA compliance documentation
- Tristan Handy and dbt community
- RGM Marketing Analytics fundamentals module
Part of the Marketing Operations series.