This page compares the tools. The Fabric how-to covers setting up TestGen, step by step.
What Fabric gives you natively
Microsoft’s data quality answer for Fabric is Purview Unified Catalog data quality. It profiles, suggests rules with AI, scores against six dimensions, and rolls scores up to data products and governance domains. Governance teams like it because it speaks their vocabulary. It also runs inside Microsoft’s own service, billed on your Azure invoice.
The limits are specific. Profiling reads a random 1-million-row sample. CSV files and column names with spaces aren’t supported. The supported sources for Fabric are lakehouses, shortcuts, and mirrored databases; Fabric Warehouse isn’t on the list. A scan takes at most 200 rules per asset, and every rule is one you attach.
Fabric itself adds materialized lake view constraints, hand-written checks that drop rows or fail a refresh. In a Fabric Warehouse, primary and foreign keys are NOT ENFORCED. Load the same key four times and you get four rows. Our Microsoft roundup counts nineteen tools in total.
Tools worth considering on Fabric
| Tool | Profiling | Generated data tests | Observability monitors | Data quality dashboards | Runs in-platform | Open source | Price basis |
|---|---|---|---|---|---|---|---|
| DataOps TestGen | 55 characteristics, 32 hygiene checks, full table | Yes, 34 test types | Freshness, volume, schema, custom metric | Scorecards, CDE score | SQL through the analytics endpoint, self-hosted app | Apache 2.0 | Free for one connection; $100/user + $100/connection per month |
| Microsoft Purview | 1M-row random sample | AI-suggested rules | Freshness rule | Scores by asset, product, domain | Microsoft-managed Spark | No | Pay-as-you-go processing units |
| Monte Carlo | Yes | Partial | Schema changes; freshness and volume opt-in | Yes | SaaS, service principal, public preview | No | Credits per monitor, on request |
| Soda | Yes (Cloud) | Partial | Metric monitoring, anomaly detection | Yes (Cloud) | Soda-hosted or self-hosted runner | Soda Core, Elastic License 2.0 | Free tier; Team $750/month |
Generated-test and profiling ratings for Monte Carlo and Soda come from DataKitchen’s comparison pages, reviewed June 2026. Vendor pages checked October 21, 2026.
Who should pick what
Pick Purview if governance owns data quality, the scores need to roll up to a domain, and a few hundred rules cover what you need. Pick Monte Carlo if you want a managed service with ML anomaly detection across Fabric and other platforms, and you can live with a preview integration. DataKitchen’s Monte Carlo comparison estimates $120K to $250K a year for 10 users and 1,000 tables, and our Purview comparison estimates $60K to $150K. Both are our estimates.
Pick TestGen if you have bronze, silver, and gold layers and nobody writing their tests. It profiles every table in full, writes the tests, and runs them through the SQL analytics endpoint, so rows stay in your tenant. The OneLake release post maps the four places to test a Fabric medallion. When gold sits in SQL Server or Azure SQL instead, TestGen connects there too.
The core profiling, testing, and monitoring engine is free and open source. Enterprise adds team access, security, Commercial Pharma tests, source-data controls for AI, and support. See the supported sources and the Azure overview.
Documentation: Connect your Data, Access Requirements, Manage Table Groups, Data quality quickstarts (no Fabric quickstart yet).