mkt-icp-research（mkt icp 研究）

ICP Research & Audience Intelligence

*Communicate Track — Step 1 of 4. Builds deep audience intelligence from real research, not assumptions.*

Inputs Required

• Product context from .agents/mkt/product-context.md (or willingness to answer product questions)

Output

• .agents/mkt/icp-research.md

Quality Gate

Before delivering, verify:

• Every VoC quote includes platform name and is from a real source (not agent-generated)
• Each persona has a habitat map with ≥3 specific channels (not just "Reddit" — which subreddit?)
• Each emotional driver is traced to ≥2 specific quotes
• Decision psychology section names specific cognitive biases and objections (not generic "they need trust")

Chain Position

Previous: none (or any skill needing audience context) | Next: mkt-imc

Before Starting

Step 0: Product Context

Check for .agents/mkt/product-context.md. If missing: INTERVIEW. Ask the user to describe their product (what, who, problem, differentiator) and save to .agents/mkt/product-context.md. Or recommend running mkt-copywriting to bootstrap it.

Required Artifacts

Optional Artifacts

Phase 1: Scope

Ask:

1. Who are we researching? (Job role, mindset, community, or use case)
2. What decisions will this inform? (Messaging? Channels? Positioning? All three?)
3. B2C or B2B? Geography? Specific segments to focus or exclude?

Phase 2: Research

You MUST use WebSearch for real discussions. Do NOT generate hypothetical quotes, invented personas, or assumed pain points.

Search Query Patterns

Use these specific queries. Adapt [topic] to the product/audience:

Multi-Platform Coverage

Search ≥4 of these categories:

VoC Quality Criteria

A good quote has:

• Emotional intensity (frustration, excitement, relief — not neutral)
• Specificity (mentions a specific tool, workflow, number, or scenario)
• Recency (within last 12 months)
• Relevance to the product category

A bad quote is:

• Generic ("I wish there was a better way")
• Old (2+ years, industry may have shifted)
• From a seller/marketer, not a real user
• A one-word reaction with no context

Collect 3 quotes per pain category. Stop when patterns repeat.

Pain Analysis (3 Levels)

MANDATORY: Habitat Mapping

For every platform where you find audience activity, document specifically:

Not "Reddit" — which subreddit. Not "LinkedIn" — which group or content type.

See references/habitat-mapping.md for density definitions.

Decision Psychology

Document:

• Trigger: What event makes them seek solutions?
• Research behavior: Where they go first, second, third
• Cognitive biases: Which are strongest? (Loss aversion? Social proof? Authority?)
• Top 3 objections: What stops them from buying? What's the psychological root of each?
• Trust signals: Who/what do they trust? What creates instant distrust?

Phase 3: Synthesize

2 Personas (max)

For each persona, provide ALL of:

• Demographics: Age range, role, industry, company size
• Pain Profile: Top 3 pains with triggers, impact, and quotes
• Decision Psychology: Research behavior, biases, objections + roots
• Habitat Map: ≥3 specific channels with density and engagement type
• VoC Quotes: 3-5 most revealing, with platform attribution

Top 3 Emotional Drivers

Core psychological motivations. Each traced to ≥2 specific quotes.

Red Flags

Language or positioning that triggers instant skepticism for this audience.

Artifact Template

---
skill: mkt-icp-research
version: 1
date: [today's date]
status: draft
---

# ICP Research

**Date:** [today]
**Skill:** mkt-icp-research
**Product:** [from product-context.md]

## Persona 1: [Name/Archetype]

**Demographics:** [Age, role, industry, company size]

### Pain Profile
1. **[Pain name]** — [description]
   - Trigger: [what causes acute pain]
   - Impact: [daily/financial/professional]
   - Quote: "[exact quote]" — [Platform, context]
   - Quote: "[exact quote]" — [Platform, context]

2. **[Pain name]** — [description]
   [Same format]

3. **[Pain name]** — [description]
   [Same format]

### Decision Psychology
- **Trigger:** [what event starts their search]
- **Research path:** [where they go 1st → 2nd → 3rd]
- **Key biases:** [which cognitive biases are strongest]
- **Objections:** (1) [objection] — root: [psychological reason]. (2) [objection] — root: [reason]. (3) [objection] — root: [reason].
- **Trust signals:** [what they trust]
- **Distrust triggers:** [what kills credibility]

### Habitat Map
| Platform | Community | Density | Engagement | Role |
|----------|-----------|---------|-----------|------|
| [Specific] | [Specific] | H/M/L | [Type] | [Role] |

## Persona 2: [Name/Archetype]
[Same format]

## Top 3 Emotional Drivers
1. **[Driver]** — [explanation]. Quotes: "[quote 1]" ([source]), "[quote 2]" ([source])
2. **[Driver]** — [explanation]. Quotes: ...
3. **[Driver]** — [explanation]. Quotes: ...

## Red Flags
- [Language/positioning that triggers skepticism and why]

## Next Step
Run `mkt-imc` to turn these insights into a communication plan.

> On re-run: rename existing artifact to `icp-research.v[N].md` and create new with incremented version.

References

• references/habitat-mapping.md — Density definitions, cross-persona analysis
• references/icp-to-imc-handoff.md — How to package outputs for IMC
• references/voice-of-customer.md — VoC collection patterns