roi-evaluation

name	roi_evaluation
description	Estimates ROI ranges for each channel in the GTM plan's channel mix — CAC, expected conversion, payback period — and flags which channels are speculative vs. evidence-backed. Triggers on "채널별 ROI", "이 전략 수익성은?", "CAC 계산".

ROI Evaluation — Channel Economics for the GTM Plan

A channel mix without unit economics is a wish list. This skill puts a number range (not a false-precision point estimate) on each channel from gtm/GTM-PLAN.md and separates evidence-backed estimates from guesses.

Process

1. Read gtm/GTM-PLAN.md and research/FINDINGS.md. Pull the channel mix and any competitor pricing/CAC signal already collected.
2. For each channel, estimate a range, not a point:
   • CAC range (low/mid/high) with the source of each bound (industry benchmark, competitor signal, first-principles calc from ad rates — name which).
   • Expected conversion rate range at each funnel stage the user can name.
   • Payback period given the pricing set in the GTM plan.
3. Classify each estimate's confidence:
   • Grounded — based on a Tier 1–2 source or the user's own historical data.
   • Benchmark — based on published category-average benchmarks (Tier 3 acceptable here, but must be named).
   • Speculative — no comparable data exists; state the assumption explicitly and what would need to be true for it to hold.
4. Compute a blended CAC and blended payback across the channel mix, weighted by planned spend allocation (ask the user for the split if not yet decided).
5. Sensitivity check. Show how payback period changes if CAC comes in at the high end vs. low end — this is the range that matters for a go/no-go call, not the midpoint alone.

Write to gtm/ROI-EVALUATION.md:

---
gtm_ref: gtm/GTM-PLAN.md
created: {ISO-8601}
---

# ROI Evaluation

## Per-Channel Economics

| Channel | CAC (low/mid/high) | Confidence | Conversion est. | Payback (mid) |
|---|---|---|---|---|
| {...} | {...} | Grounded/Benchmark/Speculative | {...} | {...} |

## Blended Economics
- Spend allocation assumed: {...}
- Blended CAC: {...}
- Blended payback: {...}

## Sensitivity
At high-end CAC, payback extends to {...}. At low-end CAC, payback shrinks
to {...}. The plan is viable if {...}.

## Speculative Assumptions (must be validated early)
1. {assumption} — invalidated by: {what evidence would kill this}
2. ...

Anti-Rationalization

• A single blended CAC number with no range hides the actual risk — always show low/mid/high.
• "Industry standard CAC is $X" without naming the industry/source is a Speculative estimate wearing a Benchmark label — reclassify honestly.
• Don't let a good blended average hide one channel with a payback period longer than the company's runway.

Evidence

Complete when every channel has a CAC range with a named confidence level, a blended payback is computed, and at least one speculative assumption is flagged with its invalidation condition.

Red Flags

• Every channel marked "Grounded" with no citations — overconfidence.
• No sensitivity analysis — only the optimistic case is shown.
• Payback period longer than the horizon stated in research/SCOPE.md with no flag raised about it.