What data governance does a small SQL Server team actually need?

A small SQL Server team needs five things: a catalog that says what each table and column means, a short list of Critical Data Elements, tests on those elements, a quality score people can see, and a named owner for each data set. That fits in a tool your team already runs, such as DataOps TestGen, without a governance program.

Key points

  • For a team of two to five people, data governance comes down to five artifacts: catalog descriptions, a Critical Data Element list, tests, a visible quality score, and an owner per data set.
  • DataOps TestGen's Data Catalog stores a description for every table and column, tags in nine categories including Business Domain and Stakeholder Group, and PII and Critical Data Element flags, and exports to Excel or an editable CSV.
  • TestGen can flag Critical Data Elements during data profiling, a table group setting that is on by default, and computes a separate quality score from CDE-flagged columns only.
  • TestGen quality scores refresh after every data profiling or test run and break down by six quality dimensions: accuracy, completeness, consistency, timeliness, uniqueness, and validity.
  • TestGen Enterprise for three users and one SQL Server connection is $400 a month at $100 per user and $100 per connection; open source is free for one user and one connection.

Most data governance advice assumes a governance office, a council, and a budget. A three-person team running SQL Server has none of those. It has a database, a few hundred tables, some Power BI reports, and people who ask whether the numbers are right.

That team needs the smallest set of habits that makes the answer checkable.

The five things

WhatWhyHow small it can be
CatalogPeople can’t trust a column they can’t interpretOne sentence per table and per column people actually use
Critical Data ElementsNot every column deserves the same attentionDozens of columns, not hundreds
TestsA rule nobody runs isn’t governanceGenerated from data profiling, plus the business rules you know
Quality scoreLeaders ask “can I trust it?”, not “which tests failed?”One score per data set, with a CDE score beside it
OwnersSomeone has to decide what correct meansOne business owner and one data team contact per data set

Leave out the rest until a real problem asks for it: glossaries, stewardship councils, policy documents, maturity assessments.

How DataOps TestGen covers it

TestGen is a data quality tool, not a governance platform. These five pieces fall out of work it already does on your SQL Server tables.

Catalog from the data profiling run

After data profiling, TestGen’s Data Catalog shows every table and column with its statistics. You add a description for each and tag it in nine categories, including Source System, Business Domain, Data Product, and Stakeholder Group. Tags set on a table group are inherited by its tables. It exports to Excel, or to a metadata CSV you can edit in bulk and import back.

Critical Data Elements and PII

Data profiling can flag likely PII columns and, through a table group setting that is on by default, likely Critical Data Elements. Review the suggestions, then set the flag to Yes, No, or Inherit by hand, in bulk, or by CSV. PII-flagged values are masked for users without PII access.

Tests

Test generation picks from 34 test types and sets thresholds from the profile. Add the rules only you know, such as “ship date is on or after order date”, with business-rule or custom SQL tests. The no-T-SQL answer walks through it.

A score people can read

TestGen’s quality scores combine data profiling and test results, refresh after every run, and break down by accuracy, completeness, consistency, timeliness, uniqueness, and validity, or by business domain and data classification. A scorecard can show a total score and a CDE score side by side. In TestGen’s color scale, 96% and above is green and below 86% is red. For putting those scores in front of Power BI users, see the SQL Server and Power BI dashboards answer.

Owners and alerts

TestGen has no separate owner field. Use the Stakeholder Group tag for the people responsible, and set notifications so failures reach them.

What it costs

Open-source TestGen is free for one user, one connection, and one project. A three-person team with one SQL Server connection on Enterprise pays $400 a month: $100 per user plus $100 per connection. Enterprise adds per-project roles, so business owners can read scores without changing tests.

Start at the SQL Server install page.

Frequently asked questions

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

Do we need a data catalog product?

Not to start. A small team needs a description for each table and column people use, a flag on the critical ones, and somewhere others can read it. TestGen's Data Catalog does that from the data profiling results you already have, and exports to Excel. Buy a dedicated catalog when lineage across many systems becomes the problem.

What is a Critical Data Element?

A Critical Data Element is a column whose errors would hurt a decision, a report, or a regulatory filing: a customer key, a revenue amount, a dispensing date. Flag a short list, usually dozens rather than hundreds, and hold those columns to stricter tests and their own quality score.

Who should own each data set?

The person who can say what correct means for it, usually in the business, with someone on the data team who fixes failures. Write both names down where people look. In TestGen, the Stakeholder Group tag holds the owners for a table group or table, and notifications route failures to the people who act on them.

When does a small team need Microsoft Purview?

When the problem is governance across many systems: lineage into Power BI, sensitivity labels, access policy, data products. Purview does those. For testing a SQL Server database, the profile, tests, and scores matter more, and Purview doesn't profile on-premises SQL Server, per Microsoft's supported-sources table updated April 2026.

Try it on your own data

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