Copyable prompt
The complete prompt
Act as a cautious paid-media analyst. Analyse only the campaign data I supply below. Do not invent, estimate or backfill any missing figure. Business objective: [OBJECTIVE] Platform and campaign type: [PLATFORM / CAMPAIGN TYPE] Reporting period and comparison period: [DATES] Market, currency and audience: [DETAILS] Conversion definitions and tracking limitations: [DETAILS] Campaign data: [PASTE REAL METRICS] Return: 1. A data-quality check listing missing fields, inconsistent definitions and limitations. 2. A table of observed performance using only supplied figures. Show formulas for any rate you calculate. 3. The three strongest patterns, with the exact supporting metrics. 4. Possible explanations clearly labelled “hypothesis”, not fact. 5. Prioritised next actions, each with the metric to monitor and the decision it should inform. 6. Questions that must be answered before making budget or sales conclusions. Do not describe clicks, leads, messaging starts or enquiries as confirmed sales. If the data cannot support a conclusion, say so plainly.
Inputs to prepare
- The reporting and comparison periods, using the same timezone and currency.
- Campaign type, audience, placements and any major changes during the period.
- Metric definitions, especially what counts as a lead, message or conversion.
- Known tracking gaps, offline outcomes and spend or revenue figures you are allowed to share.
How to read the response
Observations should cite supplied numbers. Hypotheses should sound provisional because delivery changes, creative fatigue and audience response cannot be proven from a metric table alone. Recommendations should include a monitoring metric and a decision point.
A final human check
- Recalculate rates from the original export.
- Check whether a tracking change explains an apparent performance shift.
- Separate platform-reported actions from verified business outcomes.
- Do not scale budget solely because an AI response sounds confident.
Official and reference links