Out of the laptop
SQL scripts on one person's machine become readable steps anyone can review and run.
Data & ETL engine
Catch bad loads before your dashboards do
Reconcile source and target, catch nulls, duplicates and schema drift, and check reports against the warehouse. Each check is a plain-English step.
- Reconcile every load
- Runs inside your warehouse
- Dashboard checked against source
Compare 'staging.orders' with 'dw.fact_orders' on 'order_id'
Check that row counts match for yesterday's load
Check that 'order_id' is unique in 'dw.fact_orders'
Check that the schema of 'dw.fact_orders' matches 'baseline'
Data checks the whole team can read
Runs in your warehouse
Steps compile to SQL and run where the data sits. Only results and a sample of mismatches come back.
Dashboard to source
One test reads a value off a BI report and checks it against the warehouse.
How the data engine works
What each check catches
The checks data teams write by hand, as steps.
- Reconciliation: missing, extra or changed rows between source and target
- Row counts: partial loads and silent filter changes
- Nulls and uniqueness: broken joins, duplicate loads and bad merge keys
- Schema drift: renamed, dropped or retyped columns before reports break
The work stays put
Large tables are compared where they sit, which keeps runs fast and data where your security team expects it.
- No AI model reads your rows
- Credentials come from your vault by reference
- Run checks after a load job and fail the pipeline on a broken one
The last mile of trust
A pipeline can be correct and a report still wrong because of a filter, a join or a stale extract.
Open the 'Sales dashboard' report
Remember the 'Revenue' tile as {{report.revenue}}
Check that {{report.revenue}} equals the sum of 'amount' for last month
App and data together
Tables, columns and entities sit in the App Model next to the screens that display them.
- See which data checks and UI tests a column change affects
- Data results land in Quality View beside functional results
- A release decision covers the app and the data it reports
Questions
Does an AI model read my rows?
No. Checks run as SQL inside your warehouse.
Do I need to write SQL?
No. Steps compile to SQL for you, and you can read the generated query anytime.
Can data checks run in my pipeline?
Yes. Run them from the CLI after a load job.
See one pipeline reconciled end to end
Bring a load job and the dashboard it feeds, and leave with both checked in one test.