Ideal customer profile (ICP) template, with two worked examples

This worksheet is for founders, sales leaders, RevOps and marketers writing or rewriting an ICP for a B2B company. Set aside an hour, plus an export of your closed deals. It assumes you already know why an ICP matters and how it differs from a buyer persona; if not, start with the ideal customer profile guide, which explains the concepts and how to choose between segments. This page is the worksheet itself, two completed examples and a way to check yours.

Before you start: pull the evidence

An ICP written from memory describes the customers you'd like to have. One written from data describes the customers who actually buy, stay and pay. Pull the data first:

1. Export closed deals from 12 to 36 months ago. That window gives every won customer a chance to renew. Include industry, employee count, location, deal size, sales cycle length, lead source, won or lost, and the loss reason.
2. Mark each won customer as retained or churned at 12 months, and note any that cost far more to serve than planned.
3. Pick your 10 best and 5 to 10 worst customers. Best means they renewed, were profitable, closed in a reasonable time and would recommend you. Worst means they churned, stalled or needed constant support.
4. Write down what the team believes about good customers, on a separate list. These are assumptions until the data agrees with them.

If you have fewer than 20 customers, fill the worksheet anyway and mark most fields as assumed. Your next 20 to 30 sales conversations are the test.

The ICP worksheet

Copy the four parts into a document. After each answer, write O if you observed it in customer data or A if it's an assumption, plus a few words of evidence. The fifth card turns the finished worksheet into a brief you can prospect from.

ICP name and version: [e.g. Core ICP v2]
Owner:
Last reviewed:

ONE-SENTENCE PROFILE
We serve [type of company] with [size] in [where] that [operating condition]. They usually start looking when [trigger].

FIRMOGRAPHICS (after each: O or A, and the evidence)
Industry, in words:
Industry codes (NAICS in the US, SIC in the UK):
Employee range:
Revenue range (only if it changes fit):
Geography we can sell to and support:
Business model (who they sell to):
Ownership (founder-owned, PE-backed, public, nonprofit):

OPERATING CONDITIONS (what's true about how they work when we fit; O or A)
Process or tool they've outgrown:
Structure (sites, lines, shifts, field or office teams):
Regulation or customer requirement they must meet:
What you can see from outside that shows this (careers page, website, filings):

TRIGGERS (events that start a buying process, each with a time window)
1. [event], within [months]
2.
3.

BUYING GROUP
Owns the problem (title):
Signs off (title):
Must also say yes (IT, finance, legal, quality):
Typical sales cycle:

HARD DISQUALIFIERS (any one rules the account out)
1.
2.
3.

SOFT NEGATIVES (lower the priority, don't exclude)
1.
2.

LOOKS LIKE A FIT BUT ISN'T (patterns from lost or churned accounts)
1.
2.

GOOD COMPANIES WE CAN'T SERVE WELL YET
1.

Deals used: closed between [date] and [date]. Won: [ ] Lost: [ ]
Best customers analyzed: [ ] Worst customers analyzed: [ ]

Win rate: inside ICP [ ]% / outside [ ]%
12-month retention: inside ICP [ ]% / outside [ ]%
Average annual deal value: inside ICP [ ] / outside [ ]

Assumptions to test next, and how:
1.
2.

Next review: [date, or the result that triggers a review]

Find [industry, with codes] companies in [geography] with [employee range] that [operating condition], prioritizing those where [trigger] happened in the last [N] months.

Exclude: [hard disqualifiers].
Lower priority: [soft negatives].
People to reach: [owner of the problem], [person who signs off].
Why they'd care: [one line on the problem you solve for them].

The brief is written to be handed over: a new SDR can work from it, and so can a search tool. Bebop takes a plain-language description like this, finds matching accounts and contacts and builds a playbook for each. Whatever you use, check the first 20 results against your hard disqualifiers before you trust the rest.

Must-have, signal, disqualifier or assumption?

Not every criterion carries the same weight, and mixing them up is how ICPs end up either too narrow to fill a list or too loose to guide one. Sort each criterion with the test question:

Three rules keep the sorting honest:

• Keep must-haves to three to five. More than that and you're describing one customer, not a profile.
• Make every criterion observable from outside the company or with one question on a call. "Values innovation" can't be filtered or checked. "Has no IT job titles on its team page" can.
• Take disqualifiers from lost and churned deals, not from instinct. The best disqualifier is a pattern that has already cost you money.

Separate what you've seen from what you believe

The O and A marks matter because they decide how a criterion is used. An observed criterion can be a filter. An assumed one can only be a hypothesis to test, because filtering on a wrong assumption silently removes good accounts and you never find out.

To test an assumption, run a small batch: research and contact 20 to 30 accounts that meet it, and compare their reply and meeting rates with your observed core. If the assumption holds, change the A to an O and record the evidence.

Example 1: a managed IT services provider

Illustrative example. The company, figures and evidence are fictional, to show the method.

Brightwater IT (brightwaterit.example) is a managed IT services provider (MSP) in Columbus, Ohio, with 40 clients. Most are professional-services firms, but the sales team had been chasing anyone with a server. The industry codes come from NAICS, the six-digit North American classification that groups businesses by their primary activity.

ICP name and version: Professional services core, v2
Owner: Head of Sales. Last reviewed: 14 Oct 2026

ONE-SENTENCE PROFILE
We serve accounting, law, architecture and engineering firms with 15 to 150 staff within 90 minutes of Columbus, Ohio, that have no full-time IT lead and hold confidential client data. They usually start looking when a cyber-insurance renewal or a client's security questionnaire asks for controls they can't evidence.

FIRMOGRAPHICS
Industry: accounting, legal, architecture and engineering firms [O: 31 of 40 clients]
NAICS: 5411 Legal Services; 5412 Accounting, Tax Preparation, Bookkeeping and Payroll Services; 5413 Architectural, Engineering and Related Services [O]
Employees: 15 to 150 [O: 26 of 28 best clients]
Geography: within 90 minutes' drive of Columbus, for on-site support [O: delivery cost]
Ownership: partner-owned [A]

OPERATING CONDITIONS
No full-time IT person; an office manager or part-time contractor covers it [O]
Confidential client data under professional or contractual obligations [O]
Microsoft 365 and hybrid working [O]
Visible from outside: no IT job titles on the team page [A: proxy, test it]

TRIGGERS
1. Cyber-insurance renewal within 90 days [O: cited in 11 of 18 recent wins]
2. A client sends a security questionnaire [O: 6 of 18]
3. Office move or merger in the last 6 months [A]

BUYING GROUP
Owns the problem: office manager or COO
Signs off: managing partner
Must also say yes: finance director, for contracts over $30,000 a year
Typical cycle: 5 to 9 weeks

HARD DISQUALIFIERS
1. In-house IT team of two or more
2. Under 15 staff (can't support the minimum monthly fee)
3. Wants break-fix only, no monthly agreement

SOFT NEGATIVES
1. Another MSP's contract with more than 12 months left: set a reminder
2. More than 90 minutes away: remote-only offer

LOOKS LIKE A FIT BUT ISN'T
Satellite offices of national firms: IT is decided at headquarters [O: 0 of 5 won]

EVIDENCE
Deals closed Oct 2023 to Oct 2025: 100 (27 won)
Win rate: inside ICP 37.5%, outside 17.3%
12-month retention: inside ICP 94%, outside 67%
Assumptions to test next: the "no IT job titles" proxy; office moves as a trigger

What makes this good: the triggers carry counts ("11 of 18 recent wins"), so the team knows which to prioritize. The satellite-office exclusion came from five lost deals, not a hunch. And the "no IT job titles" proxy is marked as an assumption, so it gets tested before it becomes a filter.

Example 2: a contract food-packaging supplier

Illustrative example. The company, figures and evidence are fictional, to show the method.

Larchfield Packaging (larchfieldpack.example) makes printed trays, sleeves and lidding film in the West Midlands and supplies UK food manufacturers on 12-month contracts. Its industry codes come from the condensed SIC list Companies House uses for a company's nature of business, which makes them easy to check against the register.

ICP name and version: UK chilled and bakery, v1
Owner: Commercial Director. Last reviewed: 2 Oct 2026

ONE-SENTENCE PROFILE
We supply UK food manufacturers with 50 to 500 staff that make chilled or bakery products for supermarket own-label ranges and launch new lines several times a year. They usually start looking when a range review or new listing needs short-run printed packaging faster than their current supplier can deliver.

FIRMOGRAPHICS
Industry: chilled and bakery food manufacturing [O: 19 of 24 best customers]
UK SIC: 10130 Production of meat and poultry meat products; 10710 Manufacture of bread, fresh pastry goods and cakes; 10850 Manufacture of prepared meals and dishes; 10890 Manufacture of other food products n.e.c. [O]
Employees: 50 to 500 [O]
Geography: England and Wales, next-day delivery from the West Midlands [O]
Business model: supplies UK supermarket own-label ranges [O]

OPERATING CONDITIONS
Two or more packing lines [O]
Launches or relaunches 10 or more products a year [O: best customers averaged 14]
Print runs mostly under 50,000 units [O]
Visible from outside: new-product news; NPD or packaging technologist roles on the careers page [A]

TRIGGERS
1. Retailer range review or new listing [O: 9 of 13 recent wins]
2. New NPD manager or packaging technologist in the last 6 months [O: 4 of 13]
3. Pack format review prompted by packaging EPR fees [A]

BUYING GROUP
Owns the problem: NPD manager or packaging technologist
Signs off: head of procurement
Must also say yes: technical or quality manager (food-contact compliance); finance director for contracts over £250,000
Typical cycle: 3 to 6 months, including trials

HARD DISQUALIFIERS
1. Needs glass, cans or cartons (we make trays, sleeves and film)
2. Annual packaging spend likely under £40,000
3. Buys only through unit-price e-auctions [O: 0 of 7 won at target margin]

SOFT NEGATIVES
1. Single packing line
2. Long-term agreement with a group packaging supplier

LOOKS LIKE A FIT BUT ISN'T
Foodservice-only manufacturers. Assumed to fit; 1 of 9 quoted was won, and it left after 8 months [O: now an exclusion]

EVIDENCE
Deals closed Oct 2023 to Oct 2025: 41 (16 won)
Win rate: inside ICP 13 of 24 (54%), outside 3 of 17 (18%)
12-month retention: inside ICP 12 of 13 (92%), outside 1 of 3
Assumptions to test next: EPR fees as a trigger; careers-page signals

What makes this good: the foodservice line shows an assumption being tested and reversed, which is the whole point of the O and A marks. The EPR trigger stays an assumption: under the UK's extended producer responsibility rules, large producers (more than £2 million turnover and more than 50 tonnes of packaging) pay PackUK a fee each year for household packaging, which makes it a plausible reason to review formats, but no deal has proved it yet. Note too that the outside-ICP retention (1 of 3) is too small a sample to mean much.

Validate the profile against real accounts

An ICP is a hypothesis until accounts prove it. These checks use the Brightwater numbers from Example 1.

1. Label every closed deal from your evidence window as inside or outside the profile, using only the criteria marked O.
2. Compare win rates. Brightwater won 18 of 48 opportunities inside the profile (37.5%) and 9 of 52 outside it (17.3%).
3. Compare retention. 17 of 18 clients won inside the profile were still there after 12 months (94%), against 6 of 9 outside (67%).
4. Combine them into value per opportunity. Retained annual revenue per opportunity worked = clients retained × average annual contract value ÷ opportunities. Inside: 17 × $36,000 ÷ 48 = $12,750. Outside: 6 × $30,000 ÷ 52 = about $3,460. Each in-profile opportunity was worth about 3.7 times as much, which is the case for spending research time on it.
5. Check precision and coverage against your current customers. Sort them into the two-by-two table below, then work out both rates.
6. Test the criteria on fresh accounts. Have two people independently label the same 50 prospects as fit or not. If they agree on fewer than about 45, a criterion is too vague. When Brightwater ran this, they agreed on 38 (76%), and almost every disagreement traced back to "no full-time IT lead", which you can't see from outside. That's why the proxy exists.

Brightwater's 40 current clients, sorted for step 5:

Precision is the share of matches that are good customers: 23 of 26 (88%). Coverage is the share of good customers the profile catches: 23 of 28 (82%). Look hard at the five good customers outside it: if they share something (four of Brightwater's five were dental practices), you may have a second ICP. Look at the three poor customers inside it for a missing disqualifier.

Mind the sample size. With 48 opportunities, one extra win moves the win rate by about two percentage points; with 10, it moves it by ten. Treat any segment with fewer than about 20 deals as a hint, not a finding.

Self-check: would your ICP pass a strict judge?

GTM-Bench, a June 2026 benchmark from Blackpearl Group, Bebop's parent company, scores the ICPs that AI agents write on six criteria. They work just as well as a check on one written by a person:

• Alignment: does the profile follow logically from what you sell and why?
• Specificity: does it name concrete account types, industries, sizes, geographies and roles, rather than "growing businesses"?
• Commercial relevance: does it say why these companies would need, value or buy your offer?
• Actionability: does it give enough filters, exclusions and signals to find accounts?
• Concision: is every line doing work, with no generic marketing language?
• Separation from the offer: does it describe who buys, rather than restating what you sell?

If a line in your worksheet would fit any company's ICP, cut it or make it specific.

Common mistakes

• Describing a person instead of a company. Age, education and favorite social networks belong in a buyer persona, not an ICP.
• Writing the ICP from the logos you want. If your best customers are 50-person firms, an ICP of enterprise accounts is a plan, not a profile.
• No exclusions. Without disqualifiers, every account in the industry looks like a fit and reps spend time on the ones that churn.
• Criteria nobody can check. If two reps can't independently agree whether an account fits, the criterion needs an observable proxy.
• Never revisiting it. Put a review date in part 4, and rerun the win-rate and retention comparison when results shift. For warning signs, see why your ICP is probably outdated.

Next step: pick ten accounts that match your prospecting brief and research the first three with the account profiling template.