How to Actually Understand Your ICP

The ideal customer profile decides whether every downstream go-to-market decision compounds or burns capital and time.

What is an ICP is?

An ICP is a deliberate, narrow description of the customer for whom the product is most obviously the right answer right now. It is not a description of every possible user, it is not a persona, and it is not a TAM slide.

The ICP is the specific buyer who has a high-severity, high-frequency problem your product solves, who has the budget and authority to act on it, and who is reachable through channels you can afford to run.

The persona confusion is worth killing early, because it explains most of the weak ICP work in the market. Three things separate the two.

Peronas:

• Persona is psychographic and describes what people are like whereas.

• Persona is descriptive ("this is who buys").

• Persona changes the marketing.

ICPs:

• ICP is firmographic and behavioural and describes what they are doing and what state they are in.

• ICP is exclusive ("this is who we sell to, and we say no to everyone else").

• ICP changes the company, because a real one redirects the product roadmap, the pricing, the hiring, and the channel budget

That gives you a single test. If your ICP description does not let your team say no to a prospect, it is too broad to use.

Why most ICP work fails

Most early-stage teams believe they have an ICP. What they have is a user profile: a generic description of "the customer" that never specifies who buys, who uses, who benefits, or under what conditions any of that is true. This gap causes compounding impacts on every downstream decision, so a vague ICP produces vague marketing, generic outbound, slow sales cycles, and a pipeline full of deals that close slowly or churn fast.

The data on the asymmetry showcases the importance of properly defining your ICP. Forrester found that organisations with a clearly defined ICP achieve 68 percent higher account win rates than competitors without one, and teams working from data-backed ICPs report roughly three times the sales productivity. The cost of getting this wrong is not a marginally worse funnel; it is a structurally worse one. And in a world where small wins compound, this is a major failure point.

In digital assets the failure runs deeper, because the lazy default is "the community" or "token holders". That collapses three different audiences (buyers, users, and governance participants) into a single message that lands with none of them. A treasury allocator underwriting a position, a developer integrating an SDK, and a governance voter weighing a proposal are not one customer with one job; treat them as one and the message is too generic to convert any of them.

The frameworks worth stealing from

Five frameworks define modern ICP work. Use them together rather than picking a favourite.

Lenny Rachitsky's work on identifying the ICP synthesised conversations with more than thirty B2B founders into a hard finding: almost every team that nailed its ICP described it with three attributes, drawn from a shortlist of nine:

• business type,

• company size,

• buying role,

• pain point,

• working style,

• technology used,

• price band,

• geography, and

• community presence.

Three, no more. The worked examples Lenny covered make the discipline concrete. Gong targeted US software companies using video conferencing with deal sizes between 1,000 and 100,000 dollars; Snyk targeted security-conscious Node.js developers building with JavaScript; Canva targeted social media managers and freelancers on emerging visual platforms. The principle behind all of them: aim for an ICP that is almost comically narrow, because the narrower it is, the higher the conversion rate, the lower the acquisition cost, and the cleaner the message.

Jobs to Be Done, from Clayton Christensen and Bob Moesta, reframes the question from "who is the customer?" to "what job is the customer hiring this product to do?" A job is the progress someone is trying to make in a specific context. Christensen's milkshake study is the canonical case: a chain segmented milkshake buyers by age and demographic and learned nothing, until it found the real customer was the morning commuter with a thirty-minute drive who needed a one-handed, slow-to-consume breakfast that would not stain the car. Same demographics, different job. Moesta's switching interview operationalises this by mapping the struggling moment when the old solution started failing, the push away from it and the pull toward the new option, and the anxieties and habits that almost kept the customer where they were. The output is not a persona; it is a high-resolution map of the conditions under which someone switches, and that map is the ICP.

Steve Blank and Rob Fitzpatrick supply the discipline for getting the raw signal. Blank's point is that the ICP is a hypothesis until customer interviews confirm it, and most teams skip discovery and jump straight to selling, which is the structural reason most ICPs are wrong. Fitzpatrick's Mom Test is the operating manual: ask your mother whether your idea is good and she will say yes, so the problem is never your mother, it is the question. Interviews fail when they ask hypothetical questions ("would you use this?") instead of behavioural ones ("walk me through the last time you tried to solve this"). Three rules carry most of the value: talk about the customer's life, not your idea; ask about specifics in the past, not generics in the future; talk less and listen more.

April Dunford sharpens the definition from the positioning side. The ICP is not everyone who could use the product; it is the buyer for whom the alternatives fail. That gives a useful contrarian test: if your prospect could switch to a competitor and be roughly fine, they are not your ICP, because the ICP is whoever is most disproportionately disadvantaged by the alternatives.

David Skok closes the loop with data. Tag every closed deal with attributes (size, technographic, geography, deal size, sales-cycle length, retention), and after twenty to fifty deals the segments that produced fast cycles, high contract values, and strong retention are the real ICP, while the rest were noise. The two numbers that decide whether the ICP is right are lifetime value and acquisition cost: an ICP that holds LTV/CAC above 3:1 is sustainable, whereas one running below 1:1 is not an ICP at all, just a customer who happened to buy. Lenny's empirical signal points the same way; five paying customers who all look the same is a strong early sign the ICP is real, and ten or more is hard evidence.

How to actually build your ICP

To properly build your ICP you need to run a five step process.

1. Source the raw signal first. Customer interviews are the highest-resolution input; aim for fifteen to thirty to establish a first hypothesis.

2. Sales-call recordings through Gong or Fireflies come next, because they capture how prospects describe their problem in their own words, including the exact phrases worth lifting into copy. Win/loss analysis across twenty or more closed deals is third. Product analytics surface high-engagement users who are often not the buyers, since in product-led businesses the buyer usually sits upstream of the most active user. In digital assets, on-chain behaviour is the equivalent of technographic data, and it is stronger than the B2B version because it is verifiable and real-time: wallet age, net assets and asset mix, smart-contract interaction patterns, and protocol-specific usage reveal who actually uses the product rather than who claims to. Nansen, Arkham, Dune, Etherscan, and Helius are the tools worth funding early.

3. Run the interviews properly. The single biggest failure mode is interviewing whoever is available (friends, investors, advisors) instead of people in the actual buying moment, whose answers are noise. Apply the Mom Test rules, and for any prospect who recently chose a competitor or chose to do nothing, apply the Moesta switching structure. Bring a topic map rather than a fixed script, because the most useful interviews go where you did not predict, and run two interviewers per call so one drives while the other writes.

4. Build a segment grid. Force a separation between primary, secondary, and end users, because in many products these are three different people with three different motivations and the rest of the GTM falls apart if they are merged. Capture five dimensions for each: firmographic (size, revenue, industry, geography, business model), technographic (current stack), behavioural (what they do day to day, and what triggers the search for a new solution), pain rated on severity and frequency from nought to ten, and current workarounds. The sweet spot is high-severity and high-frequency pain, and the most common workaround, which is also the hardest competitor to beat, is doing nothing.

5. Score and prioritise. Build a weighted rubric: pain severity and frequency around 40 percent, deal-size potential around 25 percent, reachability through affordable channels around 15 percent, technographic fit around 10 percent, and strategic value such as a marquee logo or expansion path around 10 percent. Score each segment; the highest scorer is the ICP and everything else is a future ICP or noise. Apollo's 2026 framing is right that the ICP is now a living data model rather than a static document, so the scoring should refresh quarterly as deals close.

Strong ICP work produces a one-page document the whole team can act on: the headline ICP statement in one sentence, the three narrow attributes, the job to be done, the top three pains scored, the current workarounds, the decision-maker, buyer, user, and end-beneficiary named separately, the channels where the ICP actually pays attention, the switching triggers, and the win and loss signals from closed-deal analysis. Every piece of marketing, sales, product, and partnership work should trace back to that page. Without it, each team rebuilds a slightly different customer in its head and the GTM splinters.

The digital asset specifics

Four adjustments separate strong digital asset ICP work from the generic version.

Separate token holders from product users. This is the most common and most expensive ICP error in the space, because the two are usually different people with different jobs. Token-holder attributes are thesis-led: belief in the protocol's long-term value, holding versus trading behaviour, governance participation, and on-chain position relative to other holdings. Product-user attributes are workflow-led: the specific job the product does, the incumbent tools, the pain map, and the switching triggers. Build a separate ICP for each, or the messaging resonates with neither.

Separate users by function. Most digital asset products serve at least one of four functionally distinct types, and each is its own ICP.

• Builders, the developers and protocol integrators, care about documentation, SDK quality, security guarantees, and gas efficiency, and you reach them through GitHub, technical Twitter, and hackathons.

• Allocators, the funds, family offices, and treasuries, care about custody, regulation, track record, and risk-adjusted returns, and you reach them through fund networks and curated events.

• Traders and retail care about narrative, speed, and returns, and you reach them through Twitter, Telegram, KOLs, and exchange listings.

• Operators, the DAO and treasury managers, care about governance tooling and multi-sig workflows, and you reach them through governance forums. A product can serve more than one, but it must build a separate ICP for each; treating builders and traders as one audience is the canonical mistake.

Treat the regulatory wrapper as a core attribute. "Retail in the US" is a fundamentally different ICP from "retail in MENA" or "retail in Singapore" because the products, partners, and channels available to each differ. For institutional ICPs the buyer's focus on compliance as much as your feature set, and compliance often decides whether they can transact at all, so a MiCA-licensed EU institution is now a different segment from an unlicensed offshore one. Most projects do not yet split these.

Avoid the "everyone in crypto" trap. "Anyone with a wallet" or "any DeFi user" produces messaging that resonates with nobody and outbound that converts at zero.

Conclusion

The ICP is the highest-leverage piece of GTM work a team can do. Get it right and every dollar spent downstream compounds; get it wrong and no amount of marketing, sales, or product execution recovers the lost leverage.

Two principles hold across every framework above. Strong ICPs are narrow: three attributes, not nine, and comically narrow beats broadly correct every time. Strong ICPs are evidence-led: customer interviews run on Mom Test and Jobs to Be Done principles, sales recordings, and win/loss data drive the definition, while internal whiteboarding does not. In digital assets, the strongest work separates token holders from product users, separates users by function, and treats on-chain behaviour and jurisdiction as core attributes rather than afterthoughts.

There is a simple test to run every quarter. Read your one-page ICP, check it against your last twenty closed deals, won and lost, and rewrite anything the data no longer supports. Most teams do not have a customer problem; they have an ICP problem, and they discover it only after they have spent the runway. Getting the definition right before the spend is the core of the go-to-market work we do at Simplicity Group, because it is the input every other decision compounds off.