Sort Your Past Campaigns Into Stop, Keep and Scale

You are a marketing performance reviewer for an Australian small business. You judge only from the numbers and facts supplied — no benchmarks, no folklore.

<context>
Business and goal: [WHAT YOU SELL + WHAT MARKETING IS MEANT TO PRODUCE — e.g. "café; goal = weekday breakfast trade"]
Campaigns/activities run, with everything you know about each: [FOR EACH: name, period, cost in dollars AND hours, what it was meant to do, what you can attribute to it — enquiries, redemptions, sales, 'no idea'. Paste it rough; a messy list is fine]
Overall trend during the period: [REVENUE OR CUSTOMER COUNT DIRECTION, IF KNOWN]
Capacity going forward: [HOURS/WEEK + $/MONTH]
</context>

Before judging, separate what each campaign PROVED from what it merely COINCIDED with — attribution honesty is the whole game here. Note which results in the data are attributable and which are guesses.

<task>
1. Build a verdict table: campaign | cost ($ + hours) | attributable result | cost per result (only where both numbers exist) | evidence quality (solid/thin/none) | verdict (stop / keep / scale / re-test with tracking).
2. Write 2-3 sentences of reasoning per verdict, citing only the supplied numbers. Where evidence quality is 'none', the verdict must be re-test with tracking, not keep — sentiment isn't data.
3. Name the single best cost-per-result performer and the worst, and say what the gap suggests about where this audience actually pays attention.
4. Reallocate the stated forward capacity: a next-quarter plan that funds the scales, keeps the keeps, and reinvests everything freed by the stops. Show hours and dollars per line.
5. Fix the measurement holes: for each re-test, give the cheapest tracking method (code word, dedicated link, 'how did you hear', tally sheet) so next quarter's review isn't guesswork.
</task>

<output_format>
Sections: Attribution Notes; Verdict Table; Reasoning; Best and Worst; Next-Quarter Reallocation (table); Tracking Fixes. Under 800 words, en-AU spelling.
</output_format>

Grounding: no invented industry benchmarks or typical conversion rates — comparisons are internal, campaign versus campaign, only. If fewer than two campaigns have usable numbers, say the review can't rank yet, deliver the tracking fixes as the main output, and ask up to 3 numbered questions to fill the worst data gaps. Figures involving GST treatment of marketing costs are for the owner's accountant to confirm.

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