Testing mode

All qa.cafe features are currently free. Features and test data can change before launch.

Database testingPrintable checklistUpdated August 2026

SQL testing checklist for QA

Review SQL evidence safely and systematically across filters, joins, NULL values, duplicates, aggregates, and transactions.

Download a clean Markdown copy and adapt it in your team workspace.

Download template No signup or tracking gate.
01

Confirm the database context

Before you run a query, confirm the database engine, environment, schema, and data refresh time. SQL syntax and behavior can differ between PostgreSQL, SQL Server, MySQL, SQLite, and other engines.

Use a read-only account for investigation when possible. Do not run copied statements until you understand their effect. Start a transaction only when the environment and team procedure support a safe rollback.

  • Confirm the engine, version, host, database, and schema.

  • Verify that the environment is approved for the investigation.

  • Use the least privileged account that can collect the evidence.

  • Record the data refresh time and relevant test-data setup.

02

Challenge result correctness

A query can run successfully and still answer the wrong question. Compare its filters and joins with the product rule. Check whether NULL values, duplicate rows, time zones, and inclusive boundaries change the result.

Inspect a small set of known records before you trust an aggregate. Reconcile counts or totals with a second method when the result supports an important decision.

  • Check every filter against the stated business rule.

  • Look for missing and duplicated rows after each join.

  • Test NULL values instead of treating them as empty strings or zero.

  • Check lower and upper boundaries, including exact boundary values.

  • Confirm time zone, precision, and date-range assumptions.

  • Review grouping keys before trusting counts, sums, or averages.

03

Preserve reproducible evidence

Save the exact query, parameters, engine, environment, and execution time with the result. A screenshot without the query or context is difficult to review and repeat.

Remove or mask personal data, credentials, and secrets before you attach evidence to a defect or shared document. Keep sensitive evidence only in an approved location.

  • Format the query so another reviewer can read it.

  • Save input parameters and the expected rule beside the result.

  • Record row counts and a small representative sample.

  • Mask sensitive fields before sharing evidence.

  • State what the query proves and what it does not prove.

Use it with your team.

Keep the parts that support a decision. Change the rest to match your product and workflow.

Download Markdown