Why is DataKitchen TestGen better than Soda?

DataOps TestGen beats Soda for most data teams because it writes your tests from a profile of 55 column characteristics, runs entirely on your infrastructure, and publishes every price. Soda leads on alert integrations, SCIM, audit logs, incident management, catalog connectors, and in-app AI. That is DataKitchen's assessment as of October 22, 2026.

Key points

  • DataOps TestGen generates tests from 34 test types after profiling 55 characteristics per column; Soda's Contract Autopilot, a preview feature, uses AI to draft data contracts that you approve in Soda Cloud.
  • TestGen 5.134.0, released October 14, 2026, added native CSV, Parquet, and JSON files and Slack, Microsoft Teams, and webhook notifications; data contracts with Open Data Contract Standard (ODCS) export are coming in November 2026.
  • Soda still leads on native Jira and ServiceNow integrations, SCIM provisioning, audit logs, incidents, prebuilt catalog connectors such as Collibra and Purview, and an in-product AI chat, per Soda's docs checked October 22, 2026.
  • Six users on three databases pay $10,800 a year for TestGen Enterprise at list price; Soda Team is $750 a month plus Soda Processing Units, and Soda publishes no per-unit rate.
  • Open-source TestGen is Apache 2.0 and includes the UI; Soda Core is under the Elastic License 2.0, and Soda's UI, MCP server, and AI features run in Soda Cloud.

Soda makes you decide what to check. TestGen profiles every table and writes most of the tests itself. If your problem is coverage across hundreds of tables, that one difference decides it, and the rest of this page is detail.

We sell TestGen, so weigh this accordingly. This page gives DataKitchen’s argument and where it runs out. The scorecard after the argument scores 36 features 0 to 3. The feature matrices are on the Soda OSS and Soda Enterprise comparison pages. Soda facts below come from soda.io/pricing and docs.soda.io, checked October 22, 2026.

Why TestGen comes out ahead

Coverage is the job most teams fail at, and hand-written checks are why. Soda Core, Great Expectations, and dbt tests all help you write tests faster. You still write each one. DataKitchen’s argument for generating 80% of tests is that the data can write the routine ones, leaving people the business rules. TestGen does that from a profile of 55 column characteristics, then adds 19 configured test types for the rules only you know.

The open-source edition is a product, not a library. You get the UI, the scorecards, the catalog, the monitors, and the MCP server under Apache 2.0. Soda Core is Elastic License 2.0, and Soda’s MCP server is for Soda Cloud customers.

Everything runs on your infrastructure. Soda’s deployment options put the scan runner in your Kubernetes cluster or Soda’s, managed from Soda Cloud.

What shipped in October 2026

Two gaps our own comparison pages listed closed in TestGen 5.134.0 on October 14, 2026. A third, data contracts, is coming in November 2026.

Files. TestGen now profiles and tests CSV, Parquet, and JSON on a local path, Amazon S3, or Azure Blob Storage.

Alerts. Notifications now go to email, Slack, Microsoft Teams, and any HTTPS webhook. Jira works through a Jira automation rule’s incoming webhook, which the alerts answer walks through.

Data contracts arrive in November 2026, with ODCS export. The receipt post explains how they will work. Soda’s contracts are older and come with AI drafting in preview.

Where Soda is better

Soda wins on enterprise plumbing. On its Enterprise plan it creates, updates, and removes users from Microsoft Entra ID with SCIM. TestGen Enterprise has single sign-on that creates a user at first sign-in and takes the admin role from your identity provider, so sign-in security is covered. What you give up is automation: an administrator assigns project roles and deletes departed users by hand. Soda Enterprise also has audit logs. TestGen gets audit logs in November 2026; until then it has none. Soda’s incidents give a failure an owner and a status; TestGen gives you issue reports.

Soda’s catalog connectors are prebuilt. TestGen reaches a catalog through its REST API, which means you write the glue. Soda also publishes orchestration guides for Airflow, Dagster, and Azure Data Factory. And Soda has an AI chat inside the product, which DataKitchen’s AI post calls useful inside its own walls. TestGen expects you to bring your own assistant over MCP.

Pick Soda if your security team requires SCIM and audit logs, if incidents route through ServiceNow, or if contracts with AI drafting are the whole program.

The price at six users

Six users on three databases pay (6 + 3) × $100 × 12 = $10,800 a year for TestGen Enterprise at list price, with unlimited tables. Soda Team is $750 a month, $9,000 a year, plus Soda Processing Units at a rate Soda doesn’t publish. Enterprise is a quote. Soda’s base is lower, though the Team plan has no SSO or role-based access. TestGen’s bill is the one you can predict. The cost answer prices other sizes, and the edition comparison shows what Enterprise adds. 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. If you’re also weighing Great Expectations, see TestGen or Great Expectations.

Install TestGen and point it at one schema. The first profiling run shows you whether the generated tests find anything.

How we built the scorecard

We built the first draft of the scorecard with Claude, Anthropic’s AI assistant, in a fresh session with no memory of our earlier work, so it started from the public pages rather than from what we believe about our own product. This was the prompt:

Build a competitive scorecard comparing the enterprise editions of Soda and DataKitchen TestGen, and publish it as a shareable web page.
RESEARCH

Read Soda's docs (docs.soda.io), its plan matrix (docs.soda.io/reference/documentation-access-and-licensing) and its live pricing page (soda.io/pricing). Confirm which column, Team or Enterprise, each feature is listed under on the pricing page.
Search the web for third-party Soda pricing (Toolworthy, Data Stack Index, Modern Data Tools, dataobservability.ai, G2). Record per-dataset rates, SPU details and dates, and flag any that conflict.
Read TestGen's docs (docs.datakitchen.io), including the test types page (cross-table and pharma tests) and the notifications page (which channels it supports), plus datakitchen.io/pricing.

We then checked every row against Soda’s pricing page and both vendors’ docs on October 22, 2026, updated TestGen to release 5.134.0, and set the scores. The judgment calls are ours. We dropped the third-party per-dataset rates for Soda from the price comparison, because Soda doesn’t publish one and the listings disagree.

Documentation: Notifications, MCP supported tools, REST API, 14 October 2026 release notes, Access requirements.

Scorecard · enterprise editions · checked October 22, 2026

Soda vs DataKitchen TestGen, feature by feature

36 features, each scored 0 to 3 from Soda’s and DataKitchen’s public docs and pricing pages. TestGen is scored as of release 5.134.0.

  • Soda
  • TestGen
  • ENTEnterprise plan only
  • 0 none · 1 partial · 2 solid · 3 best in class
Soda 76 / 108 Enterprise feature set
TestGen 81 / 108 Shipping in 5.134.0
Who leads, row by row 16 TestGen · 14 Soda · 6 tied Out of 36 features

By category

Testing and profilingmax 30
Soda 19
TestGen 25
Observabilitymax 12
Soda 11
TestGen 8
Insight and reportingmax 15
Soda 6
TestGen 15
Workflow and integrationsmax 24
Soda 21
TestGen 14
Security and deploymentmax 21
Soda 17
TestGen 13
Pricingmax 6
Soda 2
TestGen 6

Every feature

# Feature Soda Score TestGen Score
Testing and profilingSoda 19 · TestGen 25 · max 30
1 Data profiling Statistics, histograms, frequent values 2 55 characteristics per column; drives test generation and hygiene checks 3
2 Automatic test generation Contract Autopilot drafts contracts with AI, in previewENT 2 Tests generated from the profile, no AI needed 3
3 Test library breadth About 10 check families 2 34 generated and 19 configured test types 3
4 Custom SQL tests SQL checks and custom SQL monitors 3 Custom Condition, SQL Metric, and Custom Test types 3
5 Industry-specific tests None 0 12 Commercial Pharma tests: NPI, NDC, DEA, ICD, and moreENT 3
6 Data hygiene detection None 0 32 hygiene detectors 3
7 Cross-table reconciliation Reconciliation checks across sourcesENT 3 Aggregate balance and match tests, referential tests, and a documented migration-verification workflow 3
8 Group-by checks Native 2 Through aggregate tests or custom SQL 1
9 Data contracts Collaborative contracts, the core of the productENT 3 Coming November 2026: versioned contracts with ODCS export 0
10 Files Parquet and dataframes through Soda Core 2 CSV, Parquet, and JSON on a local path, Amazon S3, or Azure Blob 3
ObservabilitySoda 11 · TestGen 8 · max 12
11 Freshness, volume, and schema monitors Yes, with seasonality 3 Freshness, volume, schema, and metric monitors 3
12 ML anomaly detection Proprietary ML algorithm 3 Prediction models on monitor history 2
13 Record-level diagnostics Diagnostics Warehouse and failed-row samples 3 Drill from a failed test to the rows behind it 2
14 Lineage and impact Lineage and Impact module 2 Only through DataOps Observability, a separate product 1
Insight and reportingSoda 6 · TestGen 15 · max 15
15 Quality scoring Dataset health scores 2 Scorecards, CDE scoring, and a quality dashboard 3
16 Built-in data catalog Dataset pages 1 Catalog with metadata, hygiene, PII, results, and CDEs 3
17 PII detection Keeps PII local; does not detect it 0 Flags PII risk per column 3
18 Critical data elements Through custom attributes 1 Native CDE flags 3
19 Shareable reports Dashboards 2 Issue reports and export 3
Workflow and integrationsSoda 21 · TestGen 14 · max 24
20 Incident management Incidents with severity, lead, and status 3 Issue reports; incidents through DataOps Observability 1
21 Alert channels Slack, Teams, Jira, ServiceNow, and more 3 Email, Slack, Teams, and webhooks; Jira through a webhook 2
22 Webhooks Yes 2 Yes 2
23 Orchestration Guides for Airflow, Dagster, and Azure Data Factory 3 REST API calls from any orchestrator, plus a built-in scheduler; no prebuilt operators 2
24 Catalog integrations Prebuilt: Alation, Atlan, Collibra, data.world, Metaphor, Purview 3 Through the REST API; no prebuilt connectors 2
25 REST API Yes 2 OAuth 2.1 clients; personal access tokens on Enterprise 2
26 MCP server For Soda Cloud customers 2 112 tools in 8 groups, open source, runs under your TestGen role 3
27 In-app AI assistant Chat, root-cause analysis, and fix agentsENT 3 None; bring your own assistant over MCP 0
Security and deploymentSoda 17 · TestGen 13 · max 21
28 Single sign-on SSOENT 2 OIDC: Okta, Microsoft Entra ID, Auth0, Ping IdentityENT 2
29 User provisioning SCIM from Microsoft Entra ID creates, updates, and removes usersENT 3 No SCIM. SSO creates users at first sign-in and syncs the admin role; project roles and removals are manualENT 1
30 Role-based access Custom roles and RBACENT 3 Five project roles; the API and MCP server inherit the user’s roleENT 2
31 Audit logs YesENT 3 Coming November 2026 0
32 Self-hosting Soda-hosted or self-hosted runner; UI and metadata in Soda Cloud 2 UI, engine, and database all on your servers 3
33 Database coverage Broad, including Spark, Presto, and DuckDB 3 14 databases, including SAP HANA, OneLake, and Salesforce Data 360 2
34 Open-source depth Soda Core checks library, Elastic License 2.0 1 The full app, Apache 2.0 3
PricingSoda 2 · TestGen 6 · max 6
35 Price transparency Team price published; usage rate and Enterprise are not 1 Every price published 3
36 Cost for 6 users on 3 databases $9,000 a year plus usage on Team; SSO and RBAC need an Enterprise quote 1 $10,800 a year, unlimited tables, nothing metered 3
Total (max 108) 76 81

Pricing

Published plans

Soda Free
$0, fixed usage quota
Soda Team (unlimited users, catalog integrations)
$750/mo + usage
Soda Enterprise (SSO, RBAC, audit logs, contracts, AI)
Quote
TestGen open source (1 user, 1 connection)
$0
TestGen Enterprise
$100/user/mo + $100/connection/mo

Example: six users, three databases

Users
6
Database connections
3
Tables
Unlimited
TestGen Enterprise: (6 + 3) × $100 × 12
$10,800/yr
Soda Team base: $750 × 12
$9,000/yr + usage

Soda does not publish the rate for its usage unit, the Soda Processing Unit, so the Team figure is a floor. A team that needs SSO or RBAC is on an Enterprise quote. Warehouse compute is left out for both: each tool runs its queries in your warehouse.

  • Scores are DataKitchen’s judgment and every feature counts equally. A buyer weighting governance would rank these differently from one weighting test coverage and price.
  • A 0 means the feature is not in the vendor’s public docs. It may exist and be undocumented.
  • Plan placement for Soda follows soda.io/pricing.

Frequently asked questions

Short answers to the follow-up questions people ask next.

Does any data leave my environment?

No rows leave. DataOps TestGen runs profiling and tests as SQL inside your database, and for files it reads them in place with an embedded DuckDB engine on your own infrastructure. TestGen stores computed results plus, for display, the most frequent values and the minimum and maximum text values that profiling records. The TestGen application and its database run on your servers too.

Does TestGen support data contracts now?

Not yet. Data contracts are coming to TestGen in November 2026, with export to the Open Data Contract Standard (ODCS). Until then, TestGen's generated tests, hygiene detectors, and custom rules check the same terms a contract would state. Soda has had contracts longer, and they are on its Enterprise plan.

Is Soda cheaper than TestGen?

Sometimes, on the base fee. Soda Team is $750 a month, or $9,000 a year, plus Soda Processing Units at a rate Soda does not publish. TestGen Enterprise for six users on three databases is $10,800 a year at list price, with unlimited tables and nothing metered. Soda Team includes unlimited users, but SSO, role-based access, and audit logs are on Soda Enterprise, which is quoted. Open-source TestGen is free for one user and one database connection.

Is Soda open source?

Not under an open-source license. Soda Core uses the Elastic License 2.0, and Soda's UI, MCP server, and AI features run in Soda Cloud. TestGen is Apache 2.0. 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.

When should I choose Soda over TestGen?

Choose Soda when you need SCIM provisioning, audit logs, native ServiceNow or Jira tickets, incident management, or prebuilt connectors to Collibra, Alation, Atlan, or Purview, as of October 22, 2026. Choose it too if your team wants an AI chat inside the product. TestGen Enterprise has SSO that creates users at first sign-in, but no SCIM, so project roles and removals are manual. Its audit logs arrive in November 2026, and it connects to catalogs and AI assistants through its API and MCP server.

Try it on your own data

DataOps TestGen is free and open source. Profile a table and generate its tests this afternoon.