icp-builder

ICP Builder: Complete Methodology

The Core Principle

Your Ideal Customer Profile is not a persona slide deck that collects dust. It's an operational filter that every team uses daily — marketing uses it to target ads, SDRs use it to prioritize accounts, AEs use it to qualify deals, and CS uses it to predict churn. If your ICP is vague ("mid-market SaaS companies"), it's useless. A good ICP is specific enough that two people could independently look at the same company and agree on whether it qualifies.

---

The 4-Layer ICP Framework

Most ICPs stop at firmographics (industry + company size). That's layer 1 of 4. To build an ICP that actually predicts closed-won deals, you need all four layers.

Layer 1: Firmographic (Who they are)

These are the observable, static characteristics of the company.

Dimension	Example Criteria	Why It Matters
Industry / vertical	SaaS, FinTech, Healthcare IT	Your product may solve different problems by vertical
Company size (employees)	50-500 employees	Determines buying process complexity
Revenue range	$5M-$100M ARR	Indicates budget and sophistication
Geography	US, UK, India, DACH region	Affects compliance, language, sales motion
Growth stage	Series A-C, post-PMF	Determines urgency and budget availability
Organizational structure	Centralized vs. distributed	Affects buying committee composition
Business model	B2B, B2C, marketplace, SaaS	Determines which pain points resonate

How to extract from closed-won data: Export your last 12 months of closed-won deals from your CRM. For each deal, capture these dimensions. Look for clusters — if 70% of your wins are Series B SaaS companies with 100-300 employees, that's your firmographic sweet spot.

Layer 2: Technographic (What they use)

The tools a company uses tell you more about their sophistication, budget, and pain points than their industry does.

Dimension	Example Criteria	Why It Matters
Core platform	Salesforce vs. HubSpot vs. no CRM	Indicates maturity and integration needs
Complementary tools	Uses Snowflake + dbt + Looker	Signals data sophistication
Competitor presence	Currently uses [competitor X]	Indicates awareness of the category
Tool gaps	Has Salesforce but no enrichment tool	Signals a problem you can solve
Integration ecosystem	Heavy API usage, Zapier, custom integrations	Indicates technical sophistication
Build vs. buy tendency	In-house engineering team builds custom tools	Affects competitive positioning

How to extract from closed-won data: Use BuiltWith or your CRM's tech stack fields to profile your closed-won accounts. Look for common tools. If 80% of your customers use Salesforce, that's a technographic filter. If none of your customers have an in-house data team, "has in-house data team" might be a disqualifier.

Layer 3: Behavioral (What they do)

These are observable actions that indicate engagement, interest, or operational maturity.

Dimension	Example Criteria	Why It Matters
Content engagement	Downloaded whitepaper, attended webinar	Shows awareness of the problem
Website behavior	Visited pricing page 3+ times	Indicates active evaluation
Community participation	Active in Slack communities, Reddit, forums	Signals they're seeking solutions
Event attendance	Attended industry conferences	Indicates budget and engagement
Inbound inquiry history	Previously requested a demo	Prior interest, even if they didn't buy
Referral source	Came through partner or customer referral	Highest-quality behavioral signal

How to extract from closed-won data: Pull marketing attribution data for closed-won deals. What content did they engage with? How many touches before they booked a demo? What was the referral source? Look for patterns in the buyer journey, not just the company profile.

Layer 4: Intent / Trigger (When they're ready)

These are time-sensitive signals that indicate a company is actively looking to buy — not just a fit, but a fit right now.

Dimension	Example Criteria	Why It Matters
Hiring specific roles	Posting for RevOps, Data Engineer	Indicates investment in the area
Funding event	Raised Series B in last 90 days	Budget unlocked
Leadership change	New CRO in last 60 days	Strategic re-evaluation
Contract expiration	Competitor contract renews in Q4	Window to displace
Regulatory change	New compliance requirement	Creates urgency
Pain event	Public outage, data breach, audit failure	Immediate need

How to extract from closed-won data: For your last 20 closed-won deals, interview the AE or review the CRM notes. Ask: "What happened at this account that made them start looking?" You'll find 3-5 common triggers. These become your intent layer.

---

Reverse-Engineering ICP from Closed-Won Deals

Step 1: Export and Clean

Export your last 12-18 months of closed-won deals. Minimum 20 deals for statistical relevance, ideally 50+.

For each deal, capture:

• Company name
• Industry
• Employee count
• Revenue (estimated if not known)
• Geography
• Deal size (ACV)
• Sales cycle length
• Lead source
• Champion title
• Economic buyer title
• Key technologies used
• Trigger event (what made them start looking)
• Competitive situation (who else they evaluated)

Step 2: Segment by Deal Quality

Not all closed-won deals are equal. Segment into:

• A-deals (top 20%): Highest ACV, fastest sales cycle, lowest churn, best expansion revenue
• B-deals (middle 60%): Solid deals, average metrics
• C-deals (bottom 20%): Low ACV, long sales cycle, high churn risk, heavy support burden

Your ICP should be built from A-deals only. B-deals are acceptable targets. C-deals are accounts you should stop pursuing — they cost more to acquire and serve than they're worth.

Step 3: Find Patterns

For your A-deals, answer these questions:

Firmographic patterns:

• What industries appear most often?
• What's the employee count range? (Look for the cluster, not the outliers)
• What's the revenue range?
• What growth stage?
• What geography?

Technographic patterns:

• What CRM do they use?
• What other tools do they have?
• Do they use any competitor products?
• What's their technical sophistication?

Behavioral patterns:

• How did they find you? (Referral, inbound, outbound, event?)
• What content did they engage with before buying?
• How many stakeholders were involved?
• What was the average sales cycle?

Intent/trigger patterns:

• What event triggered the purchase?
• Was there a new hire, funding round, or competitive displacement?
• Was there a regulatory or compliance driver?

Step 4: Define Your ICP Statement

Write a one-paragraph ICP statement that combines all four layers:

Template:

Our ideal customer is a [growth stage] [industry] company with [employee range] employees and [revenue range] in annual revenue, headquartered in [geography]. They use [core technologies] and [do/don't] have [specific capability]. They are typically triggered to buy by [trigger events] and the buying process involves [champion title] and [economic buyer title] over a [timeline] sales cycle. Our average ACV with these accounts is [amount].

Worked example:

Our ideal customer is a Series A-C B2B SaaS company with 100-500 employees and $10M-$80M in ARR, headquartered in the US. They use Salesforce as their CRM and have at least one dedicated RevOps person but haven't yet built a data engineering team. They are typically triggered to buy when they hire a new VP of Sales or CRO and realize they don't have pipeline visibility. The buying process involves the RevOps Manager (champion) and VP of Sales (economic buyer) over a 30-45 day sales cycle. Our average ACV is $36K.

---

Building the Lead Scoring Model

Point-Based Scoring System

Assign points for each ICP dimension. Total points determine the lead tier.

Scoring principles:

1. Intent signals score highest (they indicate timing)
2. Firmographic fit scores second (right company type)
3. Technographic fit scores third (right tech environment)
4. Behavioral signals add bonus points
5. Negative signals subtract points (disqualification)

Score Architecture

Maximum possible score: 100 points

Category	Max Points	Weight
Firmographic fit	30	30%
Technographic fit	20	20%
Behavioral signals	20	20%
Intent signals	30	30%

Firmographic Scoring (30 points max)

Dimension	Criteria	Points
Industry	Exact ICP industry match	10
Industry	Adjacent industry	5
Industry	Non-ICP industry	0
Company size	Sweet spot (e.g., 100-500)	10
Company size	Acceptable range (e.g., 50-100 or 500-1000)	5
Company size	Outside range	0
Revenue	Sweet spot range	5
Revenue	Adjacent range	2
Geography	Target geography	5
Geography	Secondary geography	2

Technographic Scoring (20 points max)

Dimension	Criteria	Points
Core platform match	Uses target platform (e.g., Salesforce)	8
Complementary tools	Uses tools that integrate with you	5
Competitor presence	Uses a direct competitor	4
Tech sophistication	Right level (not too basic, not too advanced)	3

Behavioral Scoring (20 points max)

Dimension	Criteria	Points
Website visit (pricing page)	In last 30 days	8
Content download	In last 30 days	4
Webinar/event attendance	In last 60 days	4
Referred by customer	Any time	10
Replied to outreach	In last 90 days	6
Opened 3+ emails	In last 30 days	2

Intent Scoring (30 points max)

Dimension	Criteria	Points
Requested demo	In last 14 days	15
Relevant job posting	In last 30 days	8
Funding round	In last 60 days	8
Leadership change	In last 60 days	7
Competitor contract expiring	In next 90 days	10
Third-party intent data	Bombora/6sense surge	6

Lead Tiers

Tier	Score	Action	Owner
A-lead	70-100	Immediate outreach, multi-channel	AE or senior SDR
B-lead	50-69	Priority outreach sequence	SDR
C-lead	30-49	Nurture sequence + monitor for signals	Marketing automation
D-lead	0-29	Do not pursue	None — recycle or disqualify

---

Negative Scoring (Disqualification Signals)

These signals subtract points or automatically disqualify a lead, regardless of positive score.

Hard Disqualifiers (auto-DQ, score = 0)

• Too small: Below minimum viable company size (e.g., <10 employees for enterprise product)
• Wrong business model: B2C company for a B2B-only product
• Non-target geography: Country where you can't legally sell or support
• Already a customer: They're in your CRM as active
• Competitor employee: They work for a direct competitor
• Student/consultant: Individual using a business email for personal research
• Bankrupt/shutting down: Company is winding down operations

Soft Disqualifiers (point deductions)

Signal	Deduction	Rationale
Previous closed-lost (last 6 months)	-15	Recently evaluated and rejected
No budget authority in org	-10	Will stall at procurement
Heavy "build" culture	-10	Will try to build instead of buy
Extreme regulatory environment	-5	Long sales cycles, heavy compliance
Multiple incumbents already	-5	Hard to displace entrenched tools
Previous no-show on demo	-8	Low intent signal

---

ICP Validation

Your ICP is a hypothesis until you test it. Here's how to validate.

Test 1: Blind Sort (Internal)

Give 5 salespeople a list of 50 accounts (mix of ICP and non-ICP) without labels. Ask them to rank by "likelihood to close." Compare their rankings to your ICP scoring. If they align >70%, your ICP reflects reality. If they diverge, interview the reps to understand what they're seeing that your model isn't capturing.

Test 2: Win/Loss Correlation

Score your last 50 closed-won deals and your last 50 closed-lost deals using your new model. If the model works, closed-won deals should score significantly higher on average. If there's no meaningful difference, your scoring dimensions are wrong.

Expected results:

• Closed-won average score: 65-80
• Closed-lost average score: 30-45
• If the gap is <15 points, revisit your scoring weights

Test 3: Outbound Campaign Split

Run two outbound campaigns simultaneously:

• Campaign A: ICP-scored accounts (top tier only)
• Campaign B: Random accounts from your TAM

Compare reply rates, meeting rates, and pipeline generated. ICP-scored accounts should outperform by at least 2x on meeting rates.

Test 4: Quarterly Review

Every quarter, re-run the closed-won analysis. Your ICP should evolve as:

• You enter new markets or segments
• Your product adds capabilities
• Market conditions change
• You accumulate more closed-won data

---

Anti-Patterns

1. ICP by committee. Building your ICP in a workshop where everyone adds their pet criteria until it describes every company on earth. Build it from data first, then validate with the team.

2. Aspirational ICP. Defining your ICP as the accounts you wish you could sell to (Fortune 500) instead of the accounts you actually close (mid-market). Your ICP should reflect reality, not ambition.

3. Single-dimension ICP. "Our ICP is SaaS companies with 100-500 employees." That's one layer. Without technographic, behavioral, and intent layers, you're still spraying and praying — just at a slightly smaller list.

4. Static ICP. Building the ICP once and never updating it. Your ICP should be a living document reviewed quarterly.

5. Scoring without calibration. Assigning point values based on gut feel and never validating against outcomes. If your A-leads close at the same rate as C-leads, your scoring is broken.

6. Ignoring negative signals. Only scoring positive attributes and wondering why some high-scoring accounts never close. The disqualification layer is just as important.

---

Output Format

When building an ICP, produce two deliverables:

Deliverable 1: ICP Document

A structured document covering all four layers with specific criteria, the ICP statement paragraph, and the rationale behind each dimension. Use the template in references/icp-template.md.

Deliverable 2: Scoring Rubric

A point-based scoring model with specific criteria for each dimension, tier definitions, and disqualification rules. Use the template in references/scoring-model.md.

Both documents should be reviewed and validated quarterly using the validation tests above.