Claude Skill

ICP Refiner

Paste 5-10 won deals and 5-10 lost deals, get back patterns in good-fit vs bad-fit.

Download skill (.zip)Or download whole pack

What it does

When you have a sample of recent deals (won and lost), this skill identifies patterns in your actual ICP — company size, industry signals, tech stack indicators, role of buyer, deal velocity. Outputs a sharper ICP definition than your marketing deck has, based on what actually closes.

When to use

✓You've been selling for 6+ months and have enough deal history
✓Your stated ICP feels off — too many wrong-fit prospects entering pipeline
✓You're briefing a new SDR and need a real ICP, not aspirational

When not to use

✗You have <10 closed deals — sample is too small for pattern signal
✗You're selling something brand new where pattern hasn't emerged yet

Install

Download the .zip, then unzip into your Claude skills folder.

mkdir -p ~/.claude/skills
unzip ~/Downloads/icp-refiner.zip -d ~/.claude/skills/

# Restart Claude Code session.
# Skill is now available — Claude will use it when relevant.

SKILL.md

---
name: icp-refiner
description: Identifies real ICP patterns by analyzing won vs lost deals — company size, signals, role of buyer, deal velocity.
---

# ICP Refiner

When the user provides won and lost deals, identify patterns that separate good-fit from bad-fit prospects.

## Required input

For each deal (won AND lost):
- Company name + size (employees or revenue)
- Industry/vertical
- Buyer role (title + how senior)
- Deal size
- Days from first contact to close (or close-loss)
- Why it closed/didn't (1 sentence each)

Minimum: 5 won + 5 lost. Push back if input is smaller — patterns from <10 deals are noise.

## Analysis

Walk through these dimensions and flag where won and lost deals diverge:

### 1. Company size
Most ICP definitions are too broad here. Find the actual sweet spot.

### 2. Buyer role
Where in the org does momentum come from? Flag if won deals consistently have a different buyer than lost ones.

### 3. Velocity
Won deals are often faster. If your fastest losses are similar speed to your slowest wins, that's a signal — slow lost deals may have been disqualifiable earlier.

### 4. Trigger / signal
What was happening at the company when they engaged? Funding, hiring, leadership change, tech change. Pattern = predictive signal.

### 5. Loss reason patterns
Group lost deals by reason. If 4 of 6 losses were "no budget," budget qualification needs work earlier in the funnel. If 4 were competitor, competitive positioning is the problem.

## Output

### Refined ICP (1 paragraph)
The actual ICP based on the data. Not aspirational, descriptive of who you actually win.

### 3 disqualifiers
Patterns from your losses that mean "don't waste cycles on this prospect."

### 2 strong signals
Patterns from wins that mean "lean in, this is your zone."

### What to investigate
Where the data is suggestive but not conclusive. Worth deeper analysis.

## Constraints

- If patterns are weak, say so. Don't manufacture insight.
- Cite specific deals when claiming a pattern ("won deals 3, 5, 7 all had X")
- Push back if user wants conclusions from <10 data points

Example prompts

Once installed, try these prompts in Claude:

Here are 8 won deals and 6 lost deals from the last quarter. [paste data]. What patterns separate them?

Related prompts

Don't want to install a skill? These prompts in /prompts cover similar ground for one-shot use:

analysis prompts