Debugging wrong answers
If an answer looks off, it is usually fixable in a few checks. A wrong answer is useful feedback, not a dead end.
Quick checks
Workspace & connection
Confirm you are in the right workspace and the thread is on the intended data source.
Confirm you are in the right workspace and the thread is on the intended data source.
Timeframe & filters
Make sure the period and conditions match what you meant.
Make sure the period and conditions match what you meant.
Metric definition
Check that the measure (for example “revenue”) means what you expect here.
Check that the measure (for example “revenue”) means what you expect here.
Hidden or restricted tables
Data you cannot see, by permissions or table visibility, will not appear.
Data you cannot see, by permissions or table visibility, will not appear.
Look at how the answer was built
Open “How I found this” to see the logic, filters, data used, and any assumptions. Where it is shown, the SQL and technical details are there too. This usually reveals whether the wrong table, column, or filter was used.
If a label or description looks wrong, improving it in the AI Guide often fixes similar questions going forward.
Fix it for next time
Use the thumbs-down feedback to flag what went wrong.
Correct the description in the AI Guide, or capture the term or definition in Brief your assistant.
Rephrase with a clearer metric, timeframe, and filters, and try again.
Before you act on it
Validate important answers before making decisions. STYD aims to ground answers in your data, but you are responsible for confirming anything that matters.