# Stage 2: design questions before collecting data

**Outcome:** 5 to 15 competency questions and user stories. **Time:** about 45
minutes.

## Explain and inspect

A competency question tests what your knowledge product must answer. A user
story states a need in context: “As a local-history learner, I need to see
which source supports a date so that I separate evidence from a story.” Read
the [pack stories](personas-and-pack-stories.md) and notice that they describe
needs and limits, not screens.

## Do and check

Talk to at least one possible user or observe a real task. Include identity,
relationship, time, evidence and out-of-scope questions. Keep two questions
unseen until the final AI evaluation. For every question, write what a correct
answer must contain and what would be a serious error. Remove trivia that does
not serve the user need.

## Retrieve and reflect

Explain the difference between a competency question and a search keyword.
Record which user assumption changed after research.

Next: [research sources](03-research.md).