Four layers of semantic testing
01

Structure: fields, types, keys, relations

02

Metrics: formulas, filters, time

03

Questions: synonyms and follow-ups

04

Governance: access, versions, audit

The test object is larger than a table

Data tests verify constraints on rows and fields. Semantic tests also verify that definitions such as “gross margin” or “active customer” are selected correctly in a business question. Both are required.

Give each metric a positive case, a boundary case, and a confusing negative case.

Put equivalent questions in regression

Include formal terms, team shorthand, conversational phrasing, spelling mistakes, and follow-ups. Record expected metric, scope, time, and forbidden fields, then compare both query and result.

After a metric change, rerun the set and confirm that only expected dimensions move.

Test permissions with versions

Ask the same question as different roles and projects, then check visible data and exports. Switch semantic versions and verify that a historical result can still be replayed. BuildTable’s exact testing, versioning, and permission support requires POC confirmation.

Manage failures as defects

For each failure record question, data, version, error class, owner, and status. Classify structure, semantics, access, and interaction separately instead of patching a single prompt.

Public references

Build an AI-ready data foundation

Download BuildTable or talk with us about your data modeling scenario.

Download BuildTable