What is an ideal customer profile, and how is it different from a persona?
An ideal customer profile describes companies, not people. It answers one question: which accounts deserve your time? It covers industry, size, business model, the tools they run, the situation they are in and the problem they need solved.
A buyer persona describes the people inside those companies: their role, what they are measured on, what worries them and how they evaluate a purchase. You usually have one ICP per product and two or three personas inside it, for example the founder who signs and the operations lead who uses the product every day.
Keeping the two apart prevents a common mess: a list of "heads of marketing" at companies that could never buy, or the right companies reached through the wrong person.
| Ideal customer profile | Buyer persona | |
|---|---|---|
| Describes | A type of company | A person in a role |
| Typical fields | Industry, size, model, stack, situation, disqualifiers | Title, goals, objections, role in the purchase |
| Used for | Choosing which accounts to target | Choosing who to contact and what to say |
| Changes when | You learn who stays, pays and grows | You learn what each role cares about |
How do I build an ICP from my best customers?
The best profiles come from evidence, not a brainstorm. If you have a handful of paying customers, start there. If you have none yet, read how to get your first B2B customers and write the profile as a hypothesis you will test.
List every customer
Export customers from the last 12 to 24 months with revenue, start date, whether they renewed or expanded, length of the sales cycle and how much support they needed.
Rank them on value and effort
Mark the top fifth on revenue and retention and the bottom fifth on churn, discounts and support load. The bottom group matters as much as the top one: it tells you who to exclude.
Look for shared traits
Compare the top group on firmographics (industry, headcount, region, business model), technographics (the tools they used before you) and situation (what was happening when they bought). Situation patterns are often the most useful and the least written down.
Interview five to ten of them
Ask what triggered the search, what they tried first, who joined the decision and what nearly stopped it. Write down their exact words: they become your email copy.
Write the profile and the disqualifiers
Fill in the template below. Keep it short enough that a new hire could read a company website and say yes or no in two minutes.
Worked example: reading an ICP out of 38 customers
The company and the numbers are invented to show the method. Assume you sell a financial reporting tool that connects to small-business accounting software, and you signed 38 customers over the last 18 months. You group them by segment and add up what each group is worth and costs.
| Segment | Customers | Average annual contract | Renewed, of those due | Average sales cycle | Support load |
|---|---|---|---|---|---|
| B2B software, 20 to 200 employees | 14 | $14,000 | 11 of 12 | 35 days | Low |
| B2B services firms, 20 to 200 employees | 9 | $8,000 | 5 of 8 | 30 days | Low |
| Online retail brands | 8 | $6,000 | 2 of 7 | 20 days | High |
| Companies over 1,000 employees | 4 | $40,000 | 3 of 3 | 150 days | Very high |
| Under 20 employees | 3 | $3,000 | 1 of 3 | 10 days | Medium |
What the table says
- B2B software is the ICP. It brings the most contract value (14 × $14,000 = $196,000 a year), renews almost every time (11 of 12, about 92%) and closes in about five weeks.
- Large companies are a later bet, not the ICP. $160,000 of contract value from four accounts looks attractive, but a 150-day cycle and heavy support would swallow a small team, and four accounts are too few to judge.
- Online retail and companies under 20 people become disqualifiers. They close fast, then leave: 3 renewals out of 10 due across both groups.
- Services firms are the second ICP to test, once the first one produces meetings steadily. Their renewals (5 of 8) are decent and the cycle is short.
What the interviews add
Numbers tell you who; interviews tell you when. In this example, assume 9 of the 14 software customers say they started looking within three months of hiring their first controller, and most describe the same pain: month-end reporting rebuilt by hand in spreadsheets. That hire becomes the trigger you watch for, and their phrase becomes your subject line.
Ideal customer profile template (copy and fill in)
Copy this table into a doc or a spreadsheet and replace the examples with what you observed in your best customers, not what you hope is true. Leave a field empty rather than guess. The example column is filled in for the reporting tool from the worked example above.
| Field | Question to answer | Example of a usable answer |
|---|---|---|
| Industry or vertical | Which industries are overrepresented among your best customers? | B2B software companies |
| Company size | Which headcount or revenue band buys fastest and stays longest? | 20 to 200 employees |
| Business model | How do they make money, and does it change what they need from you? | Subscription revenue, sales-led |
| Geography and language | Where can you sell, support and invoice? | US and Canada, English |
| Tech stack | Which tools do they use that you connect to, replace or depend on? | Cloud accounting software; no data warehouse yet |
| Situation or trigger | What is usually happening at the company when they buy? | Hired a first controller in the last three months |
| Problem and its cost | What breaks without you, and what does that cost them? | Month-end reporting takes four days of spreadsheet work |
| Buying team | Who signs, who uses it, who can block it? | CFO or founder signs, controller uses, IT reviews access |
| Deal profile | Typical contract value, sales cycle and onboarding effort? | $8,000 to $20,000 a year; three to six weeks |
| Disqualifiers | What makes a company a bad fit even if it matches everything else? | Fewer than 20 people; online retail; over 1,000 employees for now |
| Evidence | Which customers is this profile based on? | 14 software customers, 11 of 12 renewals |
The one-sentence version
Compress the table into one sentence that anyone on the team can repeat: "We sell to [industry] companies with [size] in [region] that [situation], because [problem and its cost]; the [role] signs and the [role] uses it." For the example: "We sell to US and Canadian B2B software companies with 20 to 200 employees that just hired their first controller, because month-end reporting still takes them four days in spreadsheets; the CFO or founder signs and the controller uses it." If you cannot write the sentence, the table is not finished.
Where do you find the data for each field?
A field you cannot check from the outside is a field you cannot prospect with. Before you finalize the profile, confirm that each one can be looked up for a company you have never spoken to. The guide to finding B2B leads covers the sources in more depth.
| Field | Where to check it | Watch out for |
|---|---|---|
| Industry | The company's website and customer list; NAICS codes if you need a standard label | Database categories are inconsistent: a "software" company may be an agency that builds software |
| Company size | Company pages on professional networks, team and careers pages | Ranges lag behind reality; contractors distort counts |
| Business model | Pricing page (self-serve or "contact sales") and customer logos | Companies that sell to both consumers and businesses |
| Tech stack | Website technology profilers such as BuiltWith; tools named in job posts | Profilers see what runs on the public website, not back-office tools such as accounting software |
| Situation or trigger | Job posts, launches, funding news, public posts asking for help | Age: date every signal, because a six-month-old trigger has usually passed |
| Geography | Website footer, legal pages, currency on the pricing page | Headquarters and the buying team can sit in different countries |
Reach companies with a reason to buy this week
Startories finds the buying signal, verifies the decision-maker and runs the outreach until they book a call.
What does a finished ICP look like? A SaaS and an agency example
Both companies below are fictional. They show the level of detail that makes a profile usable for prospecting. For more on these two markets, see lead generation for SaaS and lead generation for marketing agencies.
Example 1: a SaaS that sells scheduling software to clinic groups
- Industry: private physiotherapy and dental clinic groups in the US.
- Size: 3 to 25 locations, 30 to 400 employees.
- Situation: opening new locations, or outgrowing a scheduling tool built for a single site.
- Problem: front desks double-book rooms and staff across locations, and managers rebuild schedules in spreadsheets every week.
- Buying team: the operations director chooses, clinic managers use it daily, the owner approves anything above a set budget.
- Deal profile: annual contract priced per location; four to eight weeks from first call to signature.
- Disqualifiers: single-location practices; groups locked into an all-in-one practice suite that already includes scheduling.
- Signals to watch: announcements of new locations, job posts for a multi-site operations role, complaints about the current scheduling tool.
Example 2: an agency that runs paid acquisition for B2B SaaS
- Industry: B2B SaaS companies selling to mid-market buyers.
- Size: 15 to 150 employees, Series A or B, with a marketing team of one to four people.
- Situation: recently funded with a growth target, or a new product line that needs pipeline quickly.
- Problem: paid campaigns bring clicks but few qualified demos, and nobody in-house owns paid search full-time.
- Buying team: the head of marketing chooses; the CEO approves retainers above a threshold.
- Deal profile: monthly retainer with a three-month minimum; two to four weeks to close.
- Disqualifiers: an in-house paid team of three or more; low-priced self-serve products where the retainer would exceed what a new customer is worth in a year.
- Signals to watch: funding announcements, a first growth marketer job post, new product launches.
How do I turn an ICP into targeting rules and signals?
A profile in a doc does nothing until it becomes rules a person or a tool can apply. Split it into three layers. Hard filters keep junk out. Scoring decides who gets attention first. Signals add timing: a company can pass every filter and still have no reason to talk to you this month. The guides to buying signals and lead scoring go deeper on the last two layers.
Write the reasons down next to every lead you contact. When replies come back as "not relevant", you can see which rule let the lead through and fix it. Startories works this way: each signal is matched to a company, checked against your ICP and scored with the reasons written out, as described in signal-based outbound.
If you prospect in the EU, collect only the data your targeting needs. Under GDPR, direct marketing can rely on legitimate interest when your interest is not outweighed by the person's rights, and the EDPB guidelines on legitimate interest explain how that balancing test works.
| Layer | What it does | Example rules |
|---|---|---|
| Hard filters | Exclude companies that cannot buy | US only; 3 or more locations; physiotherapy or dental |
| Scoring | Rank the companies that pass the filters | Plus: multi-site operations role, single-site tool in use. Minus: part of a hospital system |
| Signals | Decide when to reach out | New location announced; complaint about the current tool; operations director job post |
Is your ICP big enough for outbound?
Count the companies before you commit to a channel. Using the reporting tool again, suppose your filters return 2,400 US and Canadian B2B software companies with 20 to 200 employees (an assumption: run the count in whatever database or directory you use). At 150 new companies a month, contacting each one once takes 16 months, which is room enough. Timing narrows it further: if 3% of them show a relevant signal in a given month, also an assumption, that is about 72 companies a month with a reason to talk now. Those go first.
- A few hundred companies or fewer: automated outbound is the wrong tool. Research each account by hand, write one email at a time, and lean on referrals and events.
- A few thousand: the right range for signal-led outbound. You can afford to wait for a trigger instead of mailing everyone.
- Tens of thousands: the profile is probably too loose. Add a situation field or a disqualifier until the list describes a real group of buyers.
Which mistakes make an ICP useless?
- The wish list. The profile describes the logos you would like, not the customers who stayed. You can spot it when the evidence field is empty.
- Firmographics only. Industry and size alone return thousands of lookalikes with no reason to buy. If you cannot name a trigger, the profile is not finished.
- A persona in disguise. "Heads of marketing" is a role, not a company type. Start with the company, then choose the person.
- No disqualifiers. Without them, every borderline company gets in, and the bad-fit customers from your own data come back through the front door.
- One profile per salesperson. When each person prospects their own idea of a good fit, nobody learns anything from the results. Write one shared profile and change it together.
- Labels nobody can check. "Innovative" or "growth-minded" cannot be verified from outside. Every field should be something you can look up.
How often should I review my ideal customer profile?
Treat the first version as a hypothesis. Review it monthly during your first quarter of outbound, then every quarter, and straight away when one of these happens:
- Replies saying "not a fit" or "we already have this" cluster around one segment.
- A segment books meetings but never closes, or closes and churns within a few months.
- You ship a feature, a price change or an integration that opens a new market.
- Your best new customers do not match the profile. Find out why before you call them outliers.
How to start using your ICP for outbound
Fill in the template from your top customers this week, write the disqualifiers, and pick one or two signals that match the situation field. Build a first list of 100 to 200 companies and read a sample before sending anything: if more than a handful look wrong, tighten the filters. Then plan the rest with the outbound sales strategy guide.
When you want this to run continuously, Startories turns one ICP and one intent source into qualified, verified leads with the reasons for each, as part of its AI lead generation. The Starter plan costs $99 a month, and your first project starts with a 3-day full-access trial for $1. See plans and pricing.
Frequently asked questions
What should an ideal customer profile include?
Industry, company size, business model, geography, tech stack, the situation companies are in when they buy, the problem and its cost, the buying team, typical deal size and cycle, clear disqualifiers, and the customers the profile is based on.
What is the difference between an ICP and a buyer persona?
An ICP describes the type of company you should sell to. A buyer persona describes a person inside that company: their role, goals and objections. Use the ICP to choose accounts and the personas to choose who to contact and what to say.
Can I write an ICP before I have customers?
Yes, as a hypothesis. Base it on the problem you solve, conversations with prospects and who asked for early access. Keep it narrow, test it on a few hundred companies, and rewrite it once paying customers show you who actually buys.
How many ICPs should a startup have?
Usually one at a time per product. Several profiles split your messaging, lists and learning. Add a second ICP only when the first one produces meetings at a steady rate and you have the time to run both properly.
How many customers do I need to build an ICP from data?
There is no fixed number, but patterns become hard to trust with only a handful of customers per segment. With fewer than about ten customers in total, treat the profile as a hypothesis and lean on interviews more than on averages.
How do I know my ICP is wrong?
Watch for replies saying the email is not relevant, meetings that never become deals, and new customers who churn quickly. If a segment shows two of these, narrow it or drop it, and check whether your best recent customers still match.