Create a campaign retrospective from performance data
Author: AILesson9 min setupTested with:ChatGPTReviewed: 2026-08-28
Quick answer
Reconcile campaign results against goals, data quality, spend, delivery, and prior hypotheses. Provide: Campaign plan and hypotheses, Performance and cost data, Execution record and context. Expected result: An evidence-aware retrospective with result tables, attribution limits, lessons, and next tests.
1
Add your context
Your text stays in this browser. AILesson Prompts does not send it to a model or server.
2
Your prompt
Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt
Create a campaign retrospective from the supplied plan, data, and delivery record.
Objectives, targets, audience, dates, channels, assets, budget, hypotheses, thresholds, and guardrails:
[plan]
Metric tables, definitions, numerators, denominators, sources, windows, spend, missingness, and revisions:
[data]
Actual delivery, changes, incidents, approvals, external events, feedback, and decisions:
[delivery]
Reconcile planned versus actual before interpreting outcomes. Audit definitions, denominators, deduplication, attribution, tracking loss, overlapping channels, time windows, cost inclusion, capacity constraints, and changes made mid-campaign. Recalculate only when inputs support it and show arithmetic. Separate observed result, comparison, causal interpretation, operational explanation, stakeholder report, and hypothesis. Do not call last-touch attribution causal, infer sentiment from clicks, or label a target miss as team failure.
Report primary outcome and guardrails first; then funnel/channel contributions, audience or content segments only where pre-specified or labeled exploratory, spend and known labor, delivery quality, and qualitative signals. Compare with target and appropriate baseline; use “unknown” where attribution or denominator is unavailable. Identify what worked operationally, what evidence suggests, what remains ambiguous, and what should stop, continue, adapt, or test. Each action needs source evidence, owner or Unassigned, review date, and verification. End with data-quality fixes and a claim-to-data audit.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.
From input to outcome
A worked example
See how concrete context turns this recipe into a usable result
Actual input
Campaign plan and hypotheses
Library workshop campaign, 7-20 Sep. Primary target 36 eligible registrations for 48 seats; secondary attendance. Channels: one email 7 Sep, optional reminder 15 Sep if seats remained, posters, two partner newsletters, two Facebook posts. Budget cap $250 print. No paid media/SMS. Email-subject A/B hypothesis: concrete subject may improve unique registration-page visit rate by ≥1 point without unsubscribe/complaint harm. Maximum two emails/person.
Performance and cost data
Email delivered A 884, B 889; unique page visitors A 39, B 51; unsubscribes A 3, B 4; complaints 0/0. Seven-day window. Form: 41 eligible registrations total, 2 ineligible, 5 duplicate attempts removed; 22 chose email, 8 community center, 4 Facebook, 3 poster, 4 other as referral source. UTM last-touch visits: email 90, centers 31, Facebook 28, poster short URL 12, direct/other 44; visit totals not deduplicated across channels. Attendance 34/41 registered. Print invoice $186; labor actual communications 14h, design 5h, reviewer 4h; no hourly cost supplied.
Execution record and context
Registration opened 7 Sep. Partner 1 newsletter ran 8 Sep; Partner 2 ran 13 Sep instead of planned 9 Sep. Facebook second post was not published because capacity reached 18 Sep. Reminder email was cancelled at 41 registrations. Registration page briefly lacked Chinese interpretation detail for six hours on 7 Sep, then corrected. Three registrants emailed asking whether devices were supplied; brief did not say. No privacy/accessibility incidents reported. Communications lead needs decision on subject pattern and next campaign changes.
Example output
Primary outcome: 41 eligible registrations versus target 36, five above target and 85.4% of 48-seat capacity. Attendance was 34/41=82.9% of registrants; there is no prior attendance baseline, so do not call this improved. Guardrails: email A unsubscribes 3/884=0.34%, B 4/889=0.45%, complaints 0; no acceptance threshold was supplied, so report rather than declare equivalent. No reported privacy/accessibility incident is not proof none occurred.
A/B: A visits 39/884=4.41%; B 51/889=5.74%; observed difference +1.33 percentage points, above the 1-point operational threshold. Statistical uncertainty/test method was not supplied, and registrations cannot be joined to arm. Decision: evidence supports another pre-specified test of this subject pattern; it does not yet prove downstream registration benefit or universal adoption. Communications lead owns decision.
Attribution: 22/41 registrants self-reported email, but UTM last-touch records 90 email visits; both are descriptive and cannot establish causal channel contribution because journeys overlap and visits are not cross-channel deduplicated. Partner timing changed; Facebook reminder and second email did not run due capacity. Therefore planned-versus-actual is one email A/B, one Facebook post, posters, two partner items, $186 print, and 23 known labor hours. Spend is $64 under print cap; total campaign cost is unknown without labor rates.
Operational lessons: capacity-based cancellation correctly avoided excess frequency; continue. Fix the page/content QA checklist so language support and device provision are explicit before opening; evidence is six-hour omission and three questions. Partner 2 delay should be recorded, but no outcome impact can be isolated. Next actions: communications lead decides repeat-test by 30 Sep; content owner adds equipment field and bilingual parity check before next launch; analyst defines registration-arm joining/privacy review. Audit: all counts/rates map to supplied data; no channel causation, sentiment, cost ROI, or attendance improvement claimed.
Why this works
1
Plan-versus-actual reconciliation prevents results from being judged against a campaign that was never delivered
2
Separating observation from attribution turns performance reporting into reliable learning rather than a success story
Check the result
Are planned and actual audiences, assets, dates, budget, metrics, and thresholds reconciled
Do all rates have definitions and denominators, with attribution and tracking limits visible
Are lessons and next actions proportional to evidence rather than favorable narrative
Use it with confidence
Frequently asked questions
Practical answers about when to use this recipe, what to provide, and where human review still matters
What should I prepare before using “Create a campaign retrospective from performance data”?
For “Create a campaign retrospective from performance data,” prepare Campaign plan and hypotheses, Performance and cost data, and Execution record and context. Replace placeholders only with information you can verify. If a detail is unknown, preserve that uncertainty explicitly instead of asking the model to infer it.
When is the “Create a campaign retrospective from performance data” result not ready to use?
The result is not ready if it does not yet deliver the stated outcome—An evidence-aware retrospective with result tables, attribution limits, lessons, and next tests—from the supplied evidence, or if it relies on unresolved assumptions, missing approvals, or invented details. Use the checks as release gates: revise the source inputs or assign a named, authorized reviewer instead of polishing an unsupported output.
Which AI tools have recorded tests for “Create a campaign retrospective from performance data”?
The published test record for “Create a campaign retrospective from performance data” lists ChatGPT as of 2026-08-28. This confirms recorded runs, not guaranteed compatibility or identical results in later product versions. For another tool or version, keep every constraint visible and repeat the result checks before use.
ChatGPT