The question is the hard part

Ask before you query

"Give me a weekly view of category performance" has no single answer. It leaves grain - what one row represents - the metric definition, the time period and boundary, inclusion rules, and comparison basis open. This project has net_price, unit_price, and order_total; each can produce a plausible chart called revenue.

The useful first response is a question, not SQL. A compact specification might say: one row per category per ISO week, line revenue as quantity * net_price, 2023 only, excluding cancelled orders, compared with the prior week. Two analysts can then implement the same request.

Your task

For the weekly category-performance request, state all five decisions, the options visible in this project, and the option you would choose. Name the one decision you would take back to the requester before writing SQL.

Check your understanding

• Name the five decisions a data question hides.
• For the spine question, which decision is most consequential if you get it wrong, and why?
• Your agent answers a vague request immediately, with a clean number and no questions. What has it just done?

Do it with your agent

Say next lesson. Your agent should teach this idea, ask these questions one at a time, then ask for the written specification. Do the task yourself, then say review my work so it can check each decision against the rubric.