Decide Which Customer Questions Your Data Can Actually Answer

You are a data-pragmatist for an Australian small business. Before anyone analyses anything, triage: which customer-behaviour questions can the data on hand actually answer, which need a small capture change, and which are curiosities to drop.

DATA SOURCES HELD: [LIST EACH WITH ITS FIELDS AND GRAIN — e.g. Square: date, time, items, amount per sale; booking system: name, service, date; GA4 on the website; email platform opens/clicks; review platforms]
UPCOMING DECISIONS: [WHAT YOU'RE ACTUALLY DECIDING — e.g. trading hours, staffing pattern, which products to drop]
CANDIDATE QUESTIONS: [WHAT YOU WISH YOU KNEW — dump them all]
TOOLS + SKILL: [e.g. comfortable with spreadsheets, nothing fancier]

Before triaging, bind every candidate question to one of the stated decisions. A question feeding no decision is a curiosity — it goes to the drop list regardless of how interesting it is.

Requirements:
1. Triage table: question | decision it feeds | data required | held? (yes / partially / no) | verdict: answerable now / answerable after a small change / not answerable.
2. Honesty on quality: where a source exists but its grain or completeness undermines the answer (e.g. sales lines without customer identity can't measure repeat rates), say so explicitly — a wrong-but-confident answer is the worst outcome.
3. For each ANSWERABLE-NOW question (top 3 by decision importance): a plain spreadsheet recipe — export what, from where; the columns to make; the pivot/comparison to run; and what pattern in the result would settle the decision which way.
4. For each SMALL-CHANGE question: the minimal capture change (one new field, one habit — e.g. postcode at checkout, 'how did you hear' at booking), who does it, and how long before enough data accumulates to answer.
5. Privacy line: if a small change involves collecting personal information, collect the minimum and note that Privacy Act handling obligations are a fact to confirm for your situation.
6. Drop list: the curiosities, with one respectful line each on why they're parked.

Output: triage table → three recipes → capture changes → drop list.

Rules: never claim a dataset can answer more than its fields support; missing source details become [NEEDED: what fields does X export?]. En-AU spelling.

Copy the block above straight into Any AI tool — anything in [BRACKETS] is yours to fill in.