Validatar is a data testing and data quality platform. It runs checks against your databases, files and reports, on demand or on a schedule, and tells you when data is missing, duplicated, out of range or no longer matches its source.
The problem it solves
Bad data rarely announces itself. A nightly load finishes without an error and copies only part of the rows. A source system changes a column type, and the warehouse starts storing dates as text. The dashboards still load, and nobody finds out until a number in a meeting looks wrong.
Validatar is built to catch that first. You write down what correct means for your data, as tests, and Validatar checks it every time the data changes.
Here's the example these docs use throughout. A SQL Server order system, the Wholesale Distribution System, loads into a Snowflake warehouse, the Enterprise Data Warehouse. A Validatar test totals invoiced sales by month in both. Every month through March 2024 ties out to the cent. After that, the system has invoiced sales the warehouse doesn't: the copy stopped in April 2024, and nothing errored.
What it checks
| Quality question | Typical check |
|---|---|
| Is all the data there? | Row counts and totals that should match between a source and its target; required columns that should never be empty |
| Is it unique? | Business keys that should appear once |
| Is it valid? | Values that should fall in a range, match a pattern or follow a business rule |
| Is it consistent? | Values that should look like last week's, or like the same data in another system |
| Is it fresh? | Tables that should have been refreshed by now |
| Is the structure intact? | Columns whose names or data types changed |
A test is a query or Python script plus a rule for what passes. Validatar includes ready-made checks you can add from the Catalog without writing SQL, and you write your own in SQL or Python when the rule belongs to your business.
Where tests run
Validatar doesn't load or move data between your systems, and it never changes or fixes it. It reports what it finds. Each test sends its query to the system that holds the data:
- In your database. Queries run where the data lives. With a Snowflake or Databricks data processing engine, the comparison of the two results also runs inside your account, and only the summary comes back to Validatar.
- Through a Data Agent. For systems behind your firewall, a Data Agent installed in your network runs the queries and Python scripts. The agent reaches out to Validatar for work, so Validatar never connects into your network.
- In Validatar. For sources it can reach directly, Validatar Cloud runs the queries and Python scripts itself. On Validatar Server, your Validatar server does.
The main ideas
| Idea | What it is |
|---|---|
| Catalog | Every schema, table and column Validatar has read from your data sources, with profiles, descriptions, tests and trust scores for each one. |
| Standard test | One check on specific data: a query, what to compare it to (another query, a rule or earlier runs), and what counts as a pass. |
| Template test | One check written once and applied to many tables or columns. Validatar creates a child test for each one. A row count template on the Wholesale Distribution System creates 49 child tests, one per table. One customer found 12 critical defects across 20,000 columns this way. |
| Quick test | A ready-made template test you add from a table's or column's Tests tab in the Catalog. |
| Job and schedule | A job runs a set of tests, folders and profile sets together. A schedule runs the job at set times and emails the results. |
| Trust score | A 0 to 5 star score for each table, worked out every day from the tests that check it and how they did. |
Key Concepts and Glossary defines these and the rest of the vocabulary.
Who uses it
Validatar has two kinds of users. Creators build and maintain tests: data engineers, QA analysts and data stewards. Viewers read results, trust scores and reports, and can run tests their permissions allow, but can't create or edit tests. Admin roles on top of those control who manages users, data sources and settings. See Validatar Roles.
Validatar Cloud and Validatar Server
- Validatar Cloud: Validatar hosts, runs and upgrades it. You sign in at your instance's web address.
- Validatar Server: you install it on your own Windows or Linux server with a SQL Server or PostgreSQL database, and you choose when to upgrade.
The differences, row by row: Validatar Cloud vs. Validatar Server.
Validatar AI
Validatar AI is a chat panel inside the app that can document Catalog objects, recommend tests and create and run them for you. It's available on Validatar Cloud as a beta; contact sales to enable it. Everything it does can also be done by hand. Engineers who work in an AI coding tool can connect it to Validatar through the MCP server, on Cloud and Server from 2026.1.
Next steps
- See the stages end to end: How Validatar Works
- Pick a path for your role: Start Here
- Get a test running: Run Your First Test in 10 Minutes