Creative Decision Engine

Tell me what to do with each ad — scale it, replicate it, fix it, or kill it.

flowchart LR
  A["<b>1 · Your scored<br/>portfolio</b>"]
  B["<b>2 · Seven questions</b><br/>asked of every ad,<br/>in the same order"]
  C["<b>3 · One verb per ad</b><br/>routed to the team<br/>that owns it"]
  A --> B --> C
  style A fill:#f6f8fa,stroke:#57606a,color:#1f2328
  style B fill:#ddf4ff,stroke:#0969da,color:#1f2328
  style C fill:#dafbe1,stroke:#1a7f37,color:#1f2328

Loading

Analysis that ends in a ranked list ends too early. Somebody still has to decide what to do with ad 47, and in most teams that decision gets made from whichever number is on screen.

Every ad lands in exactly one queue. Nothing falls through without a directive.

---

Run it

1. Pick the skill you need from the table below.
2. Put its SKILL.md in your AI tool's instructions field — Claude Project, Custom GPT, Gemini Gem, or an API system prompt.
3. Put anything in its reference/ folder, and anything in this repo's reference/, into the knowledge base.

Missing a required input, they ask instead of guessing.

---

What's in here

Skill	Does
creative-scale-decision-router	The seven-step sequence. One classification and one action per ad
scale-readiness-checker	Will an early winner survive more budget, and are the clicks worth having
creative-fatigue-detector	Which ads are burning out, how fast, and when production has to start
modeled-profitability-estimator	Is it profitable when conversion data is missing or untrusted
creative-portfolio-health-checker	Is the portfolio healthy, and which team should act

They chain, but each one runs on its own.

---

How it works

1. The router asks whether the signal is there, whether it is efficient, whether it will hold, and whether the clicks are worth having
2. Scale readiness splits performance by spend rather than by date — an ad can look strong on aggregate and be weakening underneath
3. Fatigue detection reads the creative-level signal, which moves weeks before a rising CPA does
4. Portfolio health routes the intervention to the right domain before anyone briefs new creative

---

Notes

The domain routing catches the most expensive mistake in the set. A team seeing weak conversion assumes the creative is weak and briefs a replacement — but if the creative signal is strong and the media signal is weak, that brief spends a cycle proving creative cannot fix delivery.

Variant production begins at Peak, not Plateau. By the time an ad plateaus the window has already closed.

---

Files

skills/            one folder per skill — SKILL.md plus its reference files
reference/         shared reference the skills load alongside them
LICENSE            MIT

---

MIT © Raneq Barber