post-launch-analysis - SKILL.md Agent Skill

name: post-launch-analysis description: Use this skill when the user asks Claude to review live Meta campaign performance — Day 3, Day 7, or Day 14 reviews, kill/scale/iterate decisions, or any "how is my campaign doing" question. Encodes the Scale DTC Statistical Analysis Hierarchy and the AI Com Academy decision tree.

Post-Launch Analysis

Purpose: Read Meta campaign data → output kill/scale/iterate decisions per ad set, grounded in real numbers (not pattern-matched averages).

INPUT REQUIRED

Before analysis can run, Jake must provide (paste or via Meta Ads MCP):

Per ad set: spend, impressions, reach, CPM, link clicks, CTR, ATC, IC (Initiate Checkout), purchases, ROAS
Per ad: hook rate (if tracked), hold rate, video view %, thumbstop rate
Time window (last 3/7/14 days OR specific date range)
Unit economics: AOV, gross margin, break-even CPA, target ROAS

If unit economics aren't given: stop and ask. Without them, every recommendation is pattern-matched (per Hallucination Protocol). Don't proceed.

THE STATISTICAL ANALYSIS HIERARCHY (Scale DTC method)

Read metrics in this order. Earlier metrics gate later ones.

1. Hook Rate (thumbstop / 3-sec video view rate)
   ├── <25% → creative top is failing → kill or rework hook before reading any other metric
   └── ≥25% → proceed to step 2

2. Hold Rate (75% video view rate or scroll depth)
   ├── Low relative to hook rate → body of creative is losing them → rework body
   └── Healthy → proceed to step 3

3. CTR (Link Click-through Rate)
   ├── <1% → creative-to-LP message mismatch OR weak CTA → fix copy/CTA
   └── ≥1% → proceed to step 4

4. ATC Rate (Add-to-Cart per landing page visit)
   ├── <5-10% [VERIFY benchmark for category] → PDP problem, not ad → switch to shopify-store-build mode
   └── healthy → proceed to step 5

5. CVR (Purchase per ATC)
   ├── <15% [VERIFY benchmark] → checkout friction OR offer weakness → audit checkout + offer stack
   └── healthy → proceed to step 6

6. CPA vs Break-even
   ├── CPA > break-even → not profitable yet; kill or refine before scaling
   ├── CPA ≈ break-even → hold, gather more data
   └── CPA < break-even → scaling candidate (coach approval required)

The hierarchy logic: if step N is failing, don't waste time analyzing step N+1. Fix step N first.

DECISION TREE PER AD SET

Pattern	Decision
Hook rate <25% AND CTR <1%	KILL — creative isn't connecting; rework or replace
Hook rate ≥25% AND CTR ≥1% AND ATC <5%	ITERATE PDP — ad is doing its job, store is leaking
Hook rate ≥25% AND CTR ≥1% AND ATC ≥5% AND CVR <15%	ITERATE OFFER — checkout or offer mechanics breaking
All metrics in spec AND CPA > break-even	HOLD or REFINE — close to viable, watch one more day
All metrics in spec AND CPA ≈ break-even	HOLD — gather more data, don't scale prematurely
All metrics in spec AND CPA < break-even	CANDIDATE TO SCALE — pending coach approval
CPM unusually high vs benchmark	CHECK — audience overlap or competing auction; not a creative problem
Hook rate good but engagement plummets over 7 days	CREATIVE FATIGUE — rotate in new creatives, don't kill ad set

DAY 3 / 7 / 14 PROTOCOLS

Day 3 review (post-launch)

Sample size warning: at $20/day, 3 days = ~$60. Often too small for statistical confidence at the ad level.
Only kill obvious failures at Day 3 (hook rate <15%, zero ATC at significant spend, Meta delivery blockers).
Don't make scaling decisions at Day 3.

Day 7 review

Apply the hierarchy and decision tree per ad set
Identify top 1-2 winners + bottom 1-2 losers
Plan iteration for the middle (new hooks, refined targeting)
Scale only with coach approval.

Day 14 review (or campaign end)

Full synthesis
Pattern recognition across ad sets (what attribute correlates with winners?)
Insights to log to DTC Second Brain raw/performance/
Hypothesis for next campaign

OUTPUT FORMAT

Write to outputs/day-N-review-<campaign>-<date>.md:

# Day <N> Review — <Campaign Name> — <Date>

## Campaign-level summary
- Spend: $X
- Revenue: $Y
- ROAS: Z
- Break-even ROAS: <ratio>
- Status: <ON TRACK / UNDER-PERFORMING / SCALING CANDIDATE>

## Per ad set
### Ad Set A: <name>
- Hook rate: X% (gate: pass/fail)
- CTR: X% (gate: pass/fail)
- ATC rate: X% (gate: pass/fail)
- CVR: X% (gate: pass/fail)
- CPA: $X vs break-even $Y
- **Decision: KILL / ITERATE / HOLD / SCALE-CANDIDATE**
- **Reasoning:** [1-2 lines]

### Ad Set B: ...
...

## Insights to log to DTC Second Brain
- raw/performance/ : [what to drop in]
- raw/ads/ : [hook patterns observed]

## Coach approval needed for
- [Any scale recommendation]
- [Any significant structural change]

## Next action
- [Single concrete next step]

HALLUCINATION PROTOCOL — APPLIED

Every benchmark cited in this skill ("Hook rate <25% = kill") is a starting heuristic. Real thresholds depend on:

Niche (wellness vs BBQ vs apparel)
Format (video vs static)
Audience (cold vs warm)
Price point (low-ticket vs high-ticket)

When recommending a threshold, flag with [VERIFY against your account benchmarks]. Don't manufacture certainty.

ESCALATION

If Jake's CPA is 2x+ break-even after Day 7:

Recommend full creative reset, not iteration
Suggest stepping back to creative-strategy mode to rebuild the hook bank

If account-level CPM has doubled overnight:

Suggest checking for: bid changes, audience overlap, competing seasonal pressure, or pixel/CAPI issues
Don't recommend creative changes until account-level health is verified