campaign-debrief

name	campaign-debrief
description	Runs the post-campaign autopsy — goal vs actual, element-level diagnosis of what worked and what failed, keep or kill or change verdicts — and writes the portable lessons into marketing-brief.md. Use within a week of the campaign ending, while the evidence is warm.
argument-hint	[the campaign + its goal + the results you have]

Campaign Debrief — the autopsy that makes the next one cheaper

The most expensive part of a failed campaign is running the same failure again. The most expensive part of a winner is not knowing which element won. A campaign is a bundle — angle, audience, offer, creative, channel, timing — and only an element-level autopsy tells you what to carry forward and what to bury.

Inputs

• The campaign, its goal as set before launch, and every result available (spend, stage metrics, sales): $ARGUMENTS
• From marketing-brief.md if present: the plan or goal as recorded, numbers that matter (what a result is allowed to cost), and past learnings — so this debrief compounds on the last one instead of repeating it.

Do this

1. Restate the pre-launch goal. If no goal was written down before launch, that is finding number one — reconstruct the honest intent and mark it as reconstructed.
2. Goal vs actual, then the economics: the headline numbers side by side, then cost per result against what the brief says the business can afford. A campaign can hit its number and still be unaffordable.
3. Break the bundle into elements: angle, audience and targeting, offer, creative format, hook, channel, landing step, timing. Attach each element's own evidence where it exists — hook rate, click-through, conversion per stage, replies.
4. Diagnose element by element. Metric patterns point at elements. Where the data cannot isolate one, write "can't tell" — a guess recorded as a finding poisons the next campaign.
   • Stopped but did not click — the creative promised weakly.
   • Clicked but did not convert — the page or offer broke the promise.
   • Converted but unaffordable — the targeting bought the wrong buyers.
   • Bought but refunded — the offer promised what delivery could not keep.
5. Rule keep, kill, or change per element, with one line of reasoning each.
6. Extract 3–5 portable lessons — things true beyond this campaign ("this audience answers pain angles, ignores aspiration") — each citing its evidence. A lesson that only applies to this campaign is a footnote, not a lesson.
7. Write it to marketing-brief.md: the lessons and element verdicts into a dated Learnings & experiments section (create it if absent), plus any instrumentation gap that made "can't tell" verdicts necessary. The next campaign starts from here.

Output

The debrief one-pager: goal vs actual with the economics, the element table with verdicts and evidence, the portable lessons, and the single biggest thing to do differently next time. Lessons live in marketing-brief.md, not in a document nobody reopens.

Rules

• No narrative rescue. "We built awareness" is what failure says when nobody set a goal.
• Evidence over invention: every lesson cites its data. A lesson without evidence is an opinion in a lab coat.
• "Can't tell" is a legitimate verdict — record it and instrument better next time.
• Diagnose elements, not people. An autopsy that assigns blame stops being told the truth.