On this page14
- What an ICP is, and what it isn't
- The five parts of a useful ICP
- 1. Firmographics
- 2. Trigger events
- 3. Pains
- 4. Buying committee
- 5. Disqualifiers
- The ICP template (copy this table)
- Worked example 1: an AI devtool
- Worked example 2: a vertical AI company
- How to score accounts against your ICP
- How to test your ICP in two weeks
- Common ICP mistakes at seed and Series A
- Putting the ICP to work
Ideal Customer Profile Worksheet for Startups (Free Template)
An ideal customer profile (ICP) is a short, specific description of the type of company that gets the most value from your product, buys fastest and stays longest. A useful ICP template covers five things: firmographics, trigger events, pains, the buying committee and disqualifiers. If your ICP can't tell a salesperson who not to call, it isn't finished.
Below you'll find the template as a table you can copy, two worked examples (an AI devtool and a vertical AI company), a scoring sheet, and a way to test the ICP against your real pipeline. The same template is in the free ICP and positioning worksheet if you'd rather print it.
What an ICP is, and what it isn't
An ICP describes a company. A buyer persona describes a person inside that company. You need both, but founders often skip straight to personas ("Head of Engineering, 35, reads Hacker News") and end up targeting the right title at the wrong companies.
| Aspect | Ideal customer profile | Buyer persona |
|---|---|---|
| Describes | An account (company, team, practice) | An individual in the buying committee |
| Main fields | Size, industry, stack, stage, triggers | Role, goals, objections, information sources |
| Used for | Account lists, territory, ad targeting, partnerships | Messaging, content, sales scripts |
| Changes | Every 6 to 12 months as you learn | More often, by campaign |
| Failure mode | Too broad ("B2B SaaS, 50 to 5,000 employees") | Fictional detail with no evidence behind it |
Write the ICP first. Then build personas for the three or four people who show up in your deals at ICP accounts.
One more distinction worth making early: your ICP is not your total addressable market. The market is everyone who could conceivably buy. The ICP is the slice you can win now, with the product you have today, at a cost you can afford. At seed, that slice is usually much narrower than founders expect, and that's fine. A narrow ICP that converts beats a broad one that keeps your pipeline full of polite meetings that never close.
The five parts of a useful ICP
1. Firmographics
The hard filters you can pull from a database: industry, headcount, revenue band, geography, funding stage, tech stack, team size of the function you sell into. For AI products, the size of the team that uses the product (say, the number of backend engineers) usually predicts fit better than total company headcount.
2. Trigger events
Something that just happened and makes this month the right month to buy. Triggers are the most underused part of an ICP and often the most valuable. Common triggers: a new funding round, a new executive hire in the function, a compliance deadline, a migration, a public incident, a job post for the role your product replaces, a new product line.
Where to find trigger data without an expensive tool:
- Funding announcements in trade press and startup databases.
- Executive moves on LinkedIn (filter by title and "started a new position").
- Job posts for the role or skill your product replaces or supports.
- Company engineering blogs and incident post-mortems.
- Regulatory calendars and compliance deadlines in your vertical.
- Conference speaker lists, which tell you who has a new initiative to talk about.
A spreadsheet with one row per account and a "trigger seen" date column is enough to start. Outreach that opens with a real, recent trigger gets read. Outreach that opens with your product name mostly doesn't. The cold email templates for B2B founders show how to lead with one.
3. Pains
The specific, costly problems the account has, described in their words. Not "inefficiency". Something like "our two senior engineers spend Friday afternoons fixing flaky tests" or "prior authorization denials are costing us a full-time biller".
4. Buying committee
Who uses it, who champions it, who signs, who can block. In small companies this might be one or two people. In mid-market and enterprise, expect four to seven, including security, legal and finance.
5. Disqualifiers
The conditions that make a deal unlikely to close or likely to churn. Writing these down is what turns an ICP from a wish list into a tool. Examples: no budget owner for the problem, a regulated data type you can't support yet, an in-house team already building the same thing, a tech stack you don't integrate with.
The ICP template (copy this table)
| Field | Your answer | Evidence (customer, call, data) |
|---|---|---|
| Industry or vertical | ||
| Company size (headcount or revenue) | ||
| Size of the team that uses the product | ||
| Geography and language | ||
| Funding stage or ownership | ||
| Required tech stack or tools | ||
| Top 3 trigger events | ||
| Top 3 pains, in their words | ||
| Cost of the pain (time, money, risk) | ||
| Champion role | ||
| Economic buyer role | ||
| Likely blockers (security, legal, IT) | ||
| Where they look for solutions | ||
| Disqualifiers (at least 3) | ||
| Best current customers that fit |
The third column matters. Every answer should point to evidence: a named customer, a call note, a churn reason, a closed-lost deal. If you can't fill it, mark the row as a hypothesis and test it.
Worked example 1: an AI devtool
This is an illustration. Say you run an AI tool that writes and maintains integration tests for backend services.
| Field | Answer |
|---|---|
| Industry | Software companies with their own product, any vertical |
| Company size | 50 to 1,000 employees |
| Team that uses it | 10 to 80 backend engineers |
| Geography | US, UK, EU, India; English-speaking engineering teams |
| Stage | Series A to Series D, or profitable private |
| Stack | Python, Go or TypeScript services; GitHub or GitLab; existing CI pipeline |
| Triggers | New VP Engineering, post-incident review, platform migration, SOC 2 audit prep |
| Pains | Flaky tests block deploys; low coverage on legacy services; senior engineers stuck on test upkeep |
| Cost of pain | Several engineer-days a month lost to test maintenance; slower release cadence |
| Champion | Staff engineer or engineering manager for platform or developer experience |
| Economic buyer | VP Engineering or CTO |
| Blockers | Security review on code access; platform team building internal tooling |
| Where they look | GitHub, engineering blogs, Hacker News, peer Slack groups, conference talks |
| Disqualifiers | Fewer than 8 backend engineers; monorepo in an unsupported language; no CI; code can't leave a fully air-gapped environment |
Notice how much the trigger events and disqualifiers sharpen this. An outbound list built from it would skip a 400-person company with 5 engineers (disqualified) and prioritise a 120-person company that just hired a VP Engineering and posted about a production incident (two triggers). The developer tools marketing playbook covers what to do with an ICP like this once you have it.
Worked example 2: a vertical AI company
Also an illustration. Say you run an AI tool that drafts prior authorization requests for specialty medical practices in the US.
| Field | Answer |
|---|---|
| Industry | Independent specialty practices: orthopedics, cardiology, oncology |
| Company size | 10 to 150 physicians |
| Team that uses it | 3 to 30 prior auth and billing staff |
| Geography | United States, starting with 5 states with the highest payer complexity |
| Ownership | Physician-owned or private equity-backed groups |
| Stack | One of the top EHRs you integrate with; practice management system in place |
| Triggers | New practice administrator; rising denial rates; PE acquisition; payer rule changes |
| Pains | Staff spend hours per request on payer portals; denials delay procedures; turnover in billing team |
| Cost of pain | Delayed revenue, cancelled procedures, overtime and temp staffing |
| Champion | Revenue cycle manager or practice administrator |
| Economic buyer | COO, CFO or managing partner |
| Blockers | Compliance officer (HIPAA, BAA), IT for EHR access |
| Where they look | Specialty society events, practice management associations, peer referrals, trade press |
| Disqualifiers | Hospital-owned practices with central IT procurement; an EHR you don't support; fully outsourced billing |
The shape is the same as the devtool example, but the fields that matter shift. Here, ownership structure and EHR are the strongest filters, and the champion is an operations role rather than a technical one. For more on marketing into regulated verticals, see the vertical AI marketing playbook.
How to score accounts against your ICP
Once the ICP exists, turn it into a score so your team can rank accounts the same way. Keep it simple: five criteria, weighted, out of 100.
| Criterion | Weight | Score 0 to 5 | Weighted (weight x score / 5) |
|---|---|---|---|
| Firmographic fit (size, industry, stack) | 30 | ||
| Active trigger event in the last 90 days | 25 | ||
| Pain confirmed (call, post, job ad) | 20 | ||
| Champion identified | 15 | ||
| No disqualifiers present | 10 |
A rough rule I like: 75 and above is a priority account, 50 to 74 is nurture, below 50 is out. Any hard disqualifier sets the score to zero regardless of the rest.
How to test your ICP in two weeks
An ICP written in a workshop is a hypothesis. Test it against reality before you build campaigns on it.
- Pull your last 20 closed-won deals. How many match the ICP on every hard filter?
- Pull your last 20 closed-lost deals. Which disqualifiers would have caught them early?
- List your 10 best customers by retention and expansion. Do they share triggers?
- Interview 5 of those customers about what was happening the month they bought.
- Ask sales which accounts felt easy and which felt like pushing a boulder.
- Compare average deal size and cycle length for ICP versus non-ICP deals.
- Rewrite the ICP with what you learned. Date it.
If fewer than half your best customers fit the ICP, the ICP is wrong, not the customers. Rewrite it around who is actually buying and staying.
Common ICP mistakes at seed and Series A
- Making it too broad to avoid "missing out", so it filters nothing.
- Describing the dream logo instead of the customer who actually buys.
- Skipping disqualifiers because they feel negative.
- Leaving out trigger events, so outreach has no timing.
- Writing one ICP for three very different segments. Write three, then rank them.
- Never revisiting it after a pivot, a new product line or a new market.
The quiet one is the second. Founders love writing the ICP around the famous logo they hope to land. But the account that actually buys at seed is often smaller, scrappier and in more pain. Build for them first, and the famous logo tends to show up later as a reference-led deal.
Putting the ICP to work
The ICP isn't a slide for your board deck. It should change what you do on Monday.
- Build your outbound list from it, filtered by triggers.
- Rewrite homepage copy using the pains column, in the customer's words.
- Pick the three conferences and communities from the "where they look" row.
- Choose podcasts and outlets your champion and economic buyer actually read.
- Brief any agency, freelancer or PR partner with it on day one.
- Feed it into your positioning work, which is the next step after the ICP.
That last point is where most teams stall. Knowing who you're for is half the work. The other half is a sentence that makes those people feel seen, which is what the positioning statement template is for. When you need it shaped into a channel plan, the GTM planner asks for your primary buyer and builds the 90 days around them. And when the ICP is ready to turn into PR targets, it's the first thing I ask any founder for.
Your ICP is right when your sales team starts saying no to deals faster.
Want help pressure-testing your ICP before you spend on outbound or PR? Book a 30-minute teardown.