An ICP is a description of the company type most likely to buy, stay, and get real value. Here's how to build one from your actual customer data.
Rees Bayba
Founder, Astra GTM
TL;DR
An ICP (Ideal Customer Profile) is a detailed description of the company type most likely to buy your product, stay as a customer, and get real value from it. Not who you'd like to sell to -- who has actually bought and succeeded. The distinction matters. An ICP built from wishful thinking about dream customers produces campaigns that never convert. An ICP built from real customer data produces campaigns that land because they describe real companies with real problems you've solved before.
These are two different things that work together. Confusing them produces sloppy targeting.
| ICP | Buyer Persona |
|---|---|
| Describes the company | Describes the individual decision-maker |
| Firmographic + technographic + behavioral | Title, seniority, goals, objections, communication style |
| Answers: which company should we target? | Answers: who inside that company should we contact? |
| Built from customer cohort analysis | Built from sales call recordings and win/loss interviews |
| Changes slowly (product evolves) | Changes with market conditions and hiring patterns |
Define your ICP first. Then define the buyer persona within that ICP. Doing it in reverse leads to campaigns targeting the right title at the wrong company.
A complete ICP covers company characteristics across multiple dimensions. The goal is specificity, not completeness. A one-page ICP that sales reps actually use beats a 20-page document nobody reads.
Defining who NOT to target is as important as defining who to target. Churn analysis almost always reveals patterns: customers who bought and left have certain characteristics in common. Closing a bad-fit customer costs more than the deal is worth -- onboarding time, support burden, eventual churn, and the ACV going to a customer who would have stayed.
The only reliable source for building an ICP is your existing customer data. If you're pre-revenue, use your beta users, design partners, or the customers of the closest comparable products. Everything else is a hypothesis to test, not a profile to rely on.
Not market research, not competitive analysis, not founder intuition. The customers who bought, stayed, and expanded tell you exactly who your product works for. Build your ICP backward from them.
A qualitative ICP leaves too much room for judgment. When every AE interprets the ICP differently, your targeting becomes inconsistent. Turn the ICP into a numeric scoring model that any rep can apply without ambiguity.
Company size: +2 for 50-200 employees, +1 for 201-500, 0 for <50, DQ for >1,000 Funding: +1 for Series A-C, 0 for bootstrap, -1 for Series D+ or public Tech stack: +2 for tech stack match (Salesforce + HubSpot), 0 for neutral, -1 for incompatible Hiring signal: +1 if hiring in sales/revenue ops in the last 60 days Geography: +1 for North America, 0 for other Minimum score to pursue: 3+. Score below 2 = disqualify.
Most ICP failures share three root causes. Recognizing them prevents building a document that sits in a Google Drive folder and influences nothing.
What's the difference between ICP and TAM?
TAM (Total Addressable Market) is the maximum revenue you could generate if you captured every possible customer. ICP is the subset of that market where your win rate, retention, and expansion are highest. TAM is a ceiling estimate. ICP is an operational targeting tool. Campaigns built around TAM waste spend on bad-fit accounts. Campaigns built around ICP concentrate effort where you actually win.
How specific should an ICP be?
Specific enough to generate a target list of under 5,000 companies without needing a human to make judgment calls about each one. If your ICP criteria returns 50,000 companies, it's too broad. If it returns 200, it might be too narrow. A well-calibrated ICP for a mid-market B2B product typically yields 2,000-10,000 addressable companies in the primary segment.
How often should you update your ICP?
After any significant product change, major expansion into a new segment, or when close rates in your primary segment drop by more than 10 percentage points. At minimum, review it every 6 months. Treat your 10 most recent churned customers as the first input -- if their characteristics don't match your anti-ICP, your anti-ICP needs updating.
Can I have multiple ICPs?
Yes, but treat each as a separate segment with its own campaign infrastructure, copy, and AE motion. Running two ICPs with the same messaging and infrastructure produces diluted results. Many companies have a primary ICP (highest win rate, fastest close) and a secondary ICP (lower win rate but larger ACV or strategic value). The primary ICP gets the majority of outbound capacity.
What if I'm pre-revenue and have no customers?
Use the customers of the most comparable product you can find. Interview 5 founders who sell to a similar buyer. Study competitor case studies and testimonials -- they reveal which customer segments their product actually works for. Build a hypothesis ICP, run a 100-contact test campaign, and treat the reply and meeting patterns as data to refine it. Every pre-revenue ICP is a hypothesis. Ship fast, test, update.
How does ICP connect to cold email copy?
Directly. Your ICP's pain points become your email opener and problem statement. Your ICP's growth signals become your personalization angle. Your ICP's current solutions (what they're using or doing before your product) become your bridge sentence. A cold email written for a specific ICP reads completely differently from a generic cold email -- because it describes their specific situation instead of a vague category of problems.
We implement these systems end-to-end. First sends within 14 days.