Skip to content

borghei/Claude-Skills / project-management/gtm/ideal-customer-profile

ideal-customer-profile

Verified

Defines which accounts to sell to across eight dimensions with explicit exclusions, then scores real accounts against that profile. Use when defining or refining an ICP, building a qualification rubric, or diagnosing poor pipeline.

references/icp-refinement-from-data.md

View on GitHub
# ICP Refinement from Closed Customers

How to derive your real ICP from actual closed-won customers.

## 1. Why refine

Your hypothesized ICP and your real ICP differ. Always.

After 10-20 closed customers, patterns emerge that you didn't predict.
Refining ICP from data:
- Sharpens targeting
- Improves close rate
- Reduces churn (better-fit customers)
- Speeds sales cycle
- Reduces wasted CAC

## 2. The refinement process

### Step 1 — Pick the right customers
Use **top customers** (not all customers):

| Metric | Threshold |
|--------|-----------|
| Health score | top quartile |
| ARR | top 20% |
| Expansion (NRR) | > 115% |
| Tenure | > 1 year |
| Referenceability | willing to be reference |
| NPS | promoter (9-10) |

Customers who score across multiple → use these for ICP refinement.

Don't refine from:
- New customers (no signal yet)
- Churned customers (different question)
- One-off mega deals (outliers)

### Step 2 — Collect attributes per customer

For each top customer:

| Attribute | Source |
|-----------|--------|
| Industry | their site, LinkedIn |
| Size (EE) | LinkedIn, Crunchbase |
| Revenue | Crunchbase, public filings |
| Geography | LinkedIn, your CRM |
| Funding stage | Crunchbase |
| Tech stack | BuiltWith, integrations they use |
| Buyer role | your CRM (deal contact) |
| Trigger event when they bought | your CRM notes; talk to AE/CSM |
| Original lead source | your CRM |
| Average cycle | your CRM |
| Use case (what JTBD they solve) | CSM interviews |
| Champion role | your CRM |
| Decision committee | your CRM, AE/CSM |
| What alternatives they considered | discovery notes, AE |

### Step 3 — Look for patterns

Per attribute, ask: do top customers cluster?

- **Industry:** are 60%+ in one vertical?
- **Size:** narrow band? (200-500 EE vs 100-5000 EE)
- **Geography:** concentrated? (US-only vs global)
- **Tech stack:** specific tool combinations recur?
- **Buyer role:** same title most of the time?
- **Trigger event:** common pattern?
- **Champion:** specific profile?
- **Use case:** narrow set of JTBDs?

A cluster of 60%+ on a dimension = real ICP signal.

### Step 4 — Compare to stated ICP

| Dimension | Stated ICP | Actual cluster | Gap |
|-----------|-----------|----------------|-----|
| Industry | "SaaS or services" | 75% SaaS | Tighten |
| Size | 100-2000 EE | 80% in 200-500 EE | Tighten band |
| Buyer | "HR leadership" | 70% specifically VP People | Specify |
| Trigger | unspecified | 60% within 90 days of new VP People hire | Add |
| Tech stack | "any HRIS" | 80% Workday + ADP | Add as qualifier |

Refined ICP = data-validated.

### Step 5 — Validate refined ICP

Before broadcasting:

- Do AEs recognize this in their deals?
- Does marketing have addressable list?
- Are recent losses outside new ICP (i.e., not just bad luck)?
- Is the addressable market still big enough?

### Step 6 — Update artifacts

- Sales qualification rubric
- Marketing targeting (list, persona, channels)
- Product roadmap (prioritize refined ICP needs)
- Disqualification criteria
- Battle cards for adjacent segments (won't pursue)

## 3. Spotting ICP creep

Common pattern: company drifts away from ICP under sales pressure.

### Symptoms
- AE closes large deals outside ICP
- Product roadmap pulled by big-deal customers
- Churn rising in non-ICP customers
- Sales cycle lengthening
- Win rate dropping

### Diagnosis
Compare last quarter's deals vs ICP:
- % in-ICP closes
- Win rate by ICP fit
- Churn by ICP fit
- Cycle time by ICP fit

If clear gap: tighten qualification.

### Sales pushback
Sales will resist ICP discipline:
- "But this deal is huge"
- "They're a great logo"
- "We can make them happy"

Compromise:
- One-off "we'll-take-the-deal-but-not-pursue-the-segment"
- No roadmap concessions for non-ICP
- No discounting beyond standard
- No additional CS investment

## 4. ICP expansion (after beachhead)

Once beachhead is winning:

- Adjacent vertical with similar JTBD
- Adjacent size band (move up or down)
- Adjacent geography
- Adjacent use case

Each adjacency = separate ICP analysis:
- Validate it's truly adjacent (shared playbook)
- Test with 5-10 deals before committing
- Update materials per ICP

## 5. Multi-ICP companies

Some products have 2-3 distinct ICPs. Map separately:

- ICP A: mid-market HR for SaaS
- ICP B: mid-market HR for healthcare (different compliance)
- ICP C: enterprise HR (different motion entirely)

Each ICP needs:
- Distinct firmographics
- Distinct persona
- Distinct messaging
- Distinct motion (often)

Companies that pretend one ICP covers multiple = mediocre at all.

## 6. ICP discovery without customers

If you have no customers yet (or < 5):

### Use design partners
- Recruit 5-10 design partners matching hypothesized ICP
- Observe their workflows + JTBDs
- Treat their patterns as proxy ICP

### Use customer interviews
- 10-15 interviews with hypothesized ICP
- Focus on:
  - What they do today
  - What's painful
  - What they've tried
  - What they'd pay for
  - Who's in their buying committee

### Use competitor wins/losses
- Customers leaving competitors → their ICP signal
- Common attributes among competitor customers

### Use TAM / Crunchbase queries
- Build a list of 100 candidates matching hypothesized ICP
- Test outreach
- See who responds + engages

## 7. ICP refresh cadence

| Trigger | Action |
|---------|--------|
| 10+ new customers since last refresh | Refresh |
| Major product launch | Refresh |
| Major market shift | Refresh |
| Churn rising in specific segment | Diagnose; may refresh |
| Quarterly anyway | Light refresh |

Don't refresh monthly. Oscillation = no targeting discipline.

## 8. Worked example — refined ICP from data

### Original (hypothesized) ICP

```
ICP: HR teams at mid-market US companies
Buyer: HR leader
Use case: HR analytics
```

### After 25 closed customers, real patterns:

- 80% SaaS or modern services (not legacy industries)
- 85% in 200-500 EE (not 500-2000 as hypothesized)
- 75% on Workday or BambooHR + ADP/Gusto
- 70% bought within 90 days of VP People joining
- 80% had a board ask for better HR reporting
- 65% had previously tried building in Tableau/Looker first
- 90% are headquartered in US; 60% Bay Area / NYC / Austin
- Average deal $42K (range $25K-$85K)

### Refined ICP

```
ICP: US-HQ mid-market SaaS or modern-services, 200-500 EE, on Workday or
BambooHR + ADP/Gusto, with a VP People in role 0-12 months who is responding
to board ask for HR reporting depth.

Trigger: VP People hire + board ask combination.

Buyer: VP People (75%) or HR Director (25%) with CFO co-approval at $40K+.

JTBD: "Become a strategic HR partner with auditable, fast-to-produce
board-level reporting; stop being the bottleneck."

Disqualify:
- Companies < 200 EE (won't fund $25K+; HR-of-1)
- Companies > 2000 EE (need enterprise feature set we don't have yet)
- Pre-Series B (no budget for $25K HR tool)
- Heavy regulated industries (different compliance needs)
- Companies that built BI in-house (NIH bias = hard sell)

Channels: SHRM events, LinkedIn outbound to VP People in target firmographics,
HR Tech podcast guest spots, Pavilion HR community.

Average cycle: 45 days.
Average deal: $42K.
LTV target: $120K (3-year × $40K avg).
```

The refined ICP is sharper, smaller, more targetable. Marketing,
sales, product all benefit.

## Refresh checklist

Quarterly:

- [ ] Pull top-20 customers (health × revenue × tenure)
- [ ] Tag each across all 8 ICP dimensions
- [ ] Find cluster patterns
- [ ] Compare to stated ICP
- [ ] Identify refinements
- [ ] Validate with sales + marketing
- [ ] Update qualification rubric
- [ ] Update targeting materials
- [ ] Update disqualification criteria
- [ ] Communicate refined ICP to GTM team
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall