Free is easy to find in this category. The useful question is how much product you get for free, and who writes the tests.
Two 2026 events made the second half of that question sharper. GX Cloud, the commercial Great Expectations product, shut down on June 1, 2026, and GX Core carried on under Fivetran. Then Boomi bought Elementary on October 19, 2026. Both open-source cores survived. The venture-funded open source post explains why that pattern keeps repeating.
DataKitchen makes DataOps TestGen and DataOps Observability, which appear below. Check every row against the linked comparison page.
The free tools, side by side
Ratings for tools other than DataKitchen’s come from DataKitchen’s comparison pages, reviewed June 2026 (Elementary updated October 19, 2026). TestGen and Observability figures come from their product and pricing pages.
| Tool | Observability: freshness, volume, schema | Profiling | Validation: how tests get made | Deployment | License | Full product free? |
|---|---|---|---|---|---|---|
| DataOps TestGen | Yes, table monitors | Yes, 55 column characteristics | 34 test types generated from the profile, 19 you configure | Self-host; tests run as SQL in your database | Apache 2.0 | Yes, for 1 user, 1 connection, 1 project |
| DataOps Observability | Pipeline runs and timing across tools such as Airflow, dbt Core, Fivetran, and Power BI | Via TestGen, which sends it results | Rule-based alerts on run events | Self-host container | Apache 2.0 | Yes, for 1 user, 1 project |
| Soda Core | Partial | Partial | You write checks in SodaCL | Self-host library and CLI | Elastic License 2.0 since January 2026 | Partial: UI and alerting are Soda Cloud |
| Elementary | Partial | Partial | dbt tests plus rule-based anomaly and schema checks | Self-host dbt package | Apache 2.0 | Partial: static report; acquired by Boomi Oct 19, 2026 |
| GX Core | Partial | Yes | You write expectations in Python | Self-host | Apache 2.0 | Partial: GX Cloud shut down June 1, 2026 |
| dbt tests | dbt source freshness for sources | Not covered | Four built-in generic tests; you write the rest in YAML or SQL | Runs inside dbt | Open source | Not rated |
| DQOps Community | Yes | Yes | About 150 built-in checks you enable | Self-host | Apache 2.0, limited | Demo: capped at a few tables |
| Deequ | Partial | Yes | You define constraints in Scala on Spark | Self-host on Spark | Apache 2.0 | Yes |
| OpenMetadata | Partial | Yes | Rule-based test suites | Self-host | Apache 2.0 | Yes; Collate sells a hosted edition |
The dbt row comes from DataKitchen’s Microsoft data quality post, which lists dbt’s four built-in tests: unique, not_null, accepted_values, and relationships. dbt has no DataKitchen comparison page.
How to read the table
Start with the validation column. It decides your coverage more than any other. A tool that makes you write every rule caps out at the tables someone had time to cover. DataKitchen’s 80% post argues that most tests can come from the data itself, leaving people the domain rules: historical balances, transaction anomalies, trends.
Then read the last column. “Partial” there usually means the UI, the scheduler, or the alert routing lives in a paid cloud. That’s a fair business model. It also means the free edition is a library you assemble into a system, and the day the company is acquired is the day you find out which half you depend on.
Testing and observability are different jobs. A test checks values inside rows. A monitor watches the table over time. If you have to pick one first, see testing or observability first.
If you’re looking for a Monte Carlo alternative
Monte Carlo is a proprietary SaaS platform priced by table tier. DataKitchen’s Monte Carlo comparison estimates it at $120,000 to $250,000 a year for 10 users and 1,000 tables, reviewed June 2026. Monte Carlo doesn’t publish list prices, so treat that as a range.
The free tools that cover Monte Carlo’s core job, freshness, volume, and schema monitors, are TestGen, DQOps Community at demo scale, and, partially, Elementary and Soda Core. None of them is a like-for-like replacement. Monte Carlo’s field-level lineage and root-cause analysis go further than TestGen’s partial column-level lineage. For a small team, the honest test is whether you need those or just need to know a table stopped loading.
If you outgrow the free edition, the cost answer prices the next step.
Sources
| Source | What it covers |
|---|---|
| The 2026 open-source data quality and observability map | Every free tool above, including which ones ship the whole product |
| Best modern data quality tools | The scored roundup the ratings come from |