An ICP for a new market is a decision rule for who gets attention and enters outbound, what problem language you test, and which accounts you leave alone. It should describe a buying situation precisely enough to guide outbound and stay open to revision as you learn. Treating a poor response as a verdict on the offer is a trap. Talking to the wrong people about a strong idea can make you think the idea is bad, then give up.1
The output
Start by deciding what the profile must control on Monday morning: who enters outbound, what you say, and which leads you reject.
The finished profile should be precise enough to target the right personas and build a repeatable sales motion with the same messages and levers.2 An ICP describes the account or household, while a buyer persona describes the individual decision-maker inside it.3 Keep both layers in the working document, then use each for its own decision.
Write the profile as a rule a rep can apply. If two reps read it and choose different accounts, it needs more work.
Start with the buying situation
Begin with why a company enters the market. Firmographic details become useful after you understand what creates urgency and what a successful outcome looks like.
Describe the buying situation, the trigger, the job, who feels the pain, and what done looks like before deriving industry and headcount.4 Ask:
- What event makes this problem active now?
- What job becomes harder because of it?
- Who feels the pain first?
- What would that person describe as finished?
- What are they using or doing today?
Listen for a situation someone can recognize in their own week. Move on when you can explain why the account would search, who would care, and what result would make the purchase worthwhile.
Turn the situation into account and role attributes
Convert the situation into observable conditions. The profile needs enough detail to select companies and enough role detail to select the people inside them.
A useful B2B ICP includes firmographic, technographic, behavioral, pain, budget-owner, buying-cycle, and retention-success criteria.5 Make the attributes specific and measurable.6 Put a company size or revenue range in the profile,7 then add the job title or role that experiences the problem.8
Choose the three most unique and important characteristics as your initial best guess.9 They should help a rep decide whether an account belongs in the first outbound group. Keep the language close to the buying situation: a current stack, a behavior that signals active buying, a trigger, or a condition that makes the pain expensive.
Remove attributes that cannot change a decision. An attribute that cannot tell you who to contact, what to say, or whether to pass does not belong in the working version.
Choose a market wedge
A new market usually contains several plausible audiences. Choose one before writing a message so the test produces a readable signal.
Score candidate audiences on pain, willingness to pay, alternatives, and reach, then run the highest-scoring audience for a fixed period.10 Keep the other audiences available for later tests.
Six months of outreach aimed at five audiences produced zero booked calls in the described case because a message written for five audiences was written for none.11 Investigate that as a targeting problem before rewriting the whole offer. A useful wedge gives the message a single situation, a recognizable buyer, and a reason to act.
Build the first version from conversations
When customer history is thin, conversations give the profile language and conditions that a desk exercise cannot supply. Keep each conversation focused on how the person experiences the problem.
If you have zero customers, build the initial profile from five 20-minute conversations, without selling, and write it in the participants' words.12 Ask about the event that started the search, the workaround in use, the cost of leaving the problem alone, and the result they wanted. Capture repeated phrases and conditions.
AI can draft a polished profile in 30 seconds.13 Use that speed for formatting and comparison. The scarce work is making the profile narrow enough to reject leads, and that work remains yours.14
Before taking the profile into outbound, read each line as a filter. If it describes a broad market without identifying a buying situation, a pain owner, or a usable trigger, send it back for another conversation.
Score, test, and refine
The profile becomes useful when it controls account selection and produces feedback. Give the outbound test a simple scoring rule, then update the rule from what buyers do.
Score four to six attributes. One example weights industry at 25, company size at 20, tech stack at 20, geography at 15, and buying signal at 20, with scores of 70 or higher indicating high fit.15 Use the score to route accounts into the test and explain why an account was left out.
Use outbound sales data as the best signal of what is working, with more weight than leads from investors and friends.16 Look for four signs that the profile is getting closer: a significant increase in conversion rate, a significant increase in enthusiasm, a stronger desire to act now, and the nod.17 Listen for these changes across the same audience and message before changing the audience again.
Refine the ICP as products and use cases grow.18 Review it quarterly and update it meaningfully every 6 to 12 months.19 When customer history is available, compare the attributes of the last 20 closed-won customers with those of the last 20 churned customers and look for attributes that appear disproportionately in each group.20
What not to do
Review these points with the profile and the first outbound results. Avoid broad, vague ICP descriptions such as enterprises and midsize companies that use computers.21 Treating every customer and sale as equally good is a mistake.22 As the business grows, check that deals are worth the resources required to close and retain them.23
A polished firmographic card does not make a profile ready for outbound. One prompt can produce a card containing industry, size, revenue, tech stack, and personas.24 Check the initial ICP instead of assuming it is correct; most founders initially got their ICP wrong.25
Front did not think about its ICP early enough, a mistake Mathilde Collin described as one of her biggest.26 Give ICP selection comparable attention to problem selection.27