Data Quality Audit - Noble House
Data Quality Audit
Internal use — Crecera only.
Automated pipeline health checks against Bronze → Silver → Gold medallion tables. Runs on every build (daily 06:00 UTC). Results reflect current BigQuery state.
Overview
FAIL
WARN
PASS
Pipeline Freshness
Days since the latest row in each table. WARN ≥ 2 days, FAIL ≥ 4 days.
Dataset is empty - query ran successfully, but no data was returned from the database
Property Coverage (last 30 days)
Distinct location_id values seen per source. Facebook Ads and Instagram use lower thresholds due to known partial coverage.
Dataset is empty - query ran successfully, but no data was returned from the database
GA4
Dataset is empty - query ran successfully, but no data was returned from the database
Google Ads
Dataset is empty - query ran successfully, but no data was returned from the database
Facebook Ads
Note: Facebook Ads connector currently returns 0 rows — known issue tracked separately. COV-003 uses lower thresholds (WARN below 10, FAIL below 5) to avoid masking real issues.
Dataset is empty - query ran successfully, but no data was returned from the database
Google Search Console
Dataset is empty - query ran successfully, but no data was returned from the database
Google Business Profile
Dataset is empty - query ran successfully, but no data was returned from the database
Instagram Insights
Dataset is empty - query ran successfully, but no data was returned from the database
Cross-Layer Parity
Session totals compared across layers (last 7 days). A small Silver vs Bronze delta is expected — Silver filters NULL location_id rows. WARN > 2%, FAIL > 10%.
Dataset is empty - query ran successfully, but no data was returned from the database
Dev vs Prod — Gold Table Comparison
Direct comparison of Gold tables in nhhr_reporting_dev (dev) vs nhhr_reporting (prod).
Gold Table Totals
Properties with Issues
All properties match between dev and prod.