What broke, and why
Health score, last 30 days
Share of checks passing each morning. Target is 95.
Score by check type
Across all tables where the check is configured.
Tables that need attention
Lowest scores first. Each square on the layer cards is one table.
Today’s file arrivals
Files expected in the landing zone, by cut-off time.
Incidents
Every incident has the evidence, the downstream impact from lineage, the likely root cause, the fix, and the alert that was sent. Open one to see it all.
Tables
Score per check type for every monitored table. A dash means the check isn’t configured for that table, usually because it doesn’t apply (no validity rules on raw bronze, for example).
Pipelines
Run history from Databricks system tables, the Azure Data Factory monitoring API, Fabric job history and Power BI refresh history, in one place.
Runs per day, last 14 days
When checks fail
Failed checks by weekday and hour (IST), last 30 days.
Pipeline status
Duration bar shows the last run; the black tick is the 30-day p95.
How QualityLens works
The checks run inside your own Databricks or Fabric workspace as a scheduled job, with read-only access to your data. Only results leave it: counts, scores and pass or fail, never your rows.
Where data comes from
- SAP ECC / S/4HANA
- POS, WMS and vendor files
- Shopify, Salesforce, APIs
- GST portal, bank files
Checks run here, every load
- Databricks job or Fabric notebook, read-only
- Checks configured per table in one YAML file
- Results written to
dq.check_results - Lineage from Unity Catalog / Purview for impact
Who hears about it
- Teams or Slack channel
- WhatsApp to on-call for critical
- Email digest at 8:30 a.m.
- This dashboard, refreshed each run
Nine check types
Each one is a tested SQL template. Onboarding a table means adding a few lines of config, not writing new code.
Who gets told, and how fast
Default routing. We agree the channels and owners with you during onboarding.
| Severity | Example | Channel | Alert within |
|---|---|---|---|
| Critical | A dashboard the business opens at 9 a.m. will be wrong | Teams + WhatsApp to on-call | 5 minutes |
| High | A silver or gold table is wrong or late | Teams channel, tagged owner | 15 minutes |
| Medium | Drift that will hurt if it continues | Teams channel | 1 hour |
| Low | Housekeeping, slow trends | Morning email digest | Next 8:30 a.m. |
Find out what’s breaking in your pipelines
Book a free 30-minute review. We look at your last 30 days of pipeline runs and the five tables your main dashboard depends on, and show you what a morning report would have caught.