Founders ask Collective 54 this 16 times in our records. It usually surfaces when win rates are inconsistent, when outbound is not landing, or when a founder is deciding whether to niche down further.
An ideal client profile that is not anchored to a specific, urgent, funded problem is a demographic exercise. It will produce a list of companies and no useful guidance about which of them will buy.
Greg Alexander test for whether a problem is worth building a firm around is direct. Can you state it plainly enough that people outside your industry understand it. Does it exist across more than one industry, across company sizes, across geographies. Are clients paying to solve it today, and have they been for years. Are the consequences of ignoring it severe. Is there a trigger event that puts a client in the market, and do they work to a deadline once they have the problem.
The underlying rule is to be in the painkiller business rather than the vitamin business. Alexander describes a would-be founder who came to Capital 54 with an elegant solution to demand curve modeling for information security companies. Asked to describe the problem, he explained that clients were unaware they had one and he intended to bring it to their attention. Asked how many had it, he said all of them. Asked whether it was urgent, he said it was not, but it should be. Three strikes, and none of them were about the quality of his expertise.
If your ICP work starts with who you would like to sell to, you will make a version of that mistake. Start with what is already hurting and who is already paying.
You know who your client is if you can produce two things: a demographic profile and a psychographic profile.
The demographic half is the one everyone builds. Industry, company size, geography, title, role, seniority, budget authority. It is necessary and it is not sufficient, because it describes a company rather than a buyer.
The psychographic half is what most firms skip. Wants, needs, goals, values, attitudes, priorities, challenges, and the emotions the buyer brings to the decision. Services are bought by people. Alexander describes a professional translator with a genuinely faster delivery model who, asked to name his client, said software publishers expanding internationally. That is a description of a company, and a thin one. The questions that actually matter were different. Why was this person put in charge of choosing a translation service. Are they confident they can pick the right one, or is this their first time. Is their career at risk if they choose badly, and are they worried about how their boss will react.
That second set is where your positioning, your proposal and your first meeting come from. The demographic profile tells you who to call. The psychographic profile tells you what to say.
A practical test: can you state the client personal and professional goals, the obstacles standing in the way of each, the objections they are likely to raise, and their top priorities. If not, the work to do is primary research. Interview the people themselves. Secondhand market research will not get you there.
An ICP can be perfectly accurate and commercially useless. Size of the prize has to justify the effort, and it takes roughly as much effort to pursue a small market as a large one.
The arithmetic is simple: number of clients multiplied by revenue per engagement is the addressable market, and the levers are penetration rate and engagement size rather than target count, because most markets already contain thousands of names. Boutiques typically penetrate their target markets by less than ten percent, so build the model on modest penetration rather than a heroic share.
The variable founders most often miss is reachability. Alexander is candid that he got this wrong himself. SBI served the head of sales at business-to-business companies, and counted roughly 12.5 million such companies as its market. That was wrong twice. The service was bought by two people rather than one, because chief marketing officers hired the firm frequently, which doubled the number. And many of those targets were unreachable, because their gatekeepers had gatekeepers. What was left was the early-adopter community, and the firm let those people identify themselves by publishing, starting with a book and extending into a blog, podcast, video and print magazine, which produced about 250,000 self-identified early adopters.
That is the useful move. Rather than assuming a market is reachable, build a mechanism that lets the right buyers raise their hands, then define the ICP from who actually responds.
Clients turn to boutiques for specialization, and they pay more for specialists than generalists. There are five dimensions to specialize along: industry, function, segment, problem and geography. A firm that helps product managers at enterprise software companies in Silicon Valley move to the cloud is specialized on all five, and its yield is high because it can charge more. This is the same work as defining an ICP, seen from the pricing side, and three to five dimensions is the range that separates you from the generalist.
Two further filters are worth applying. The first is call point, meaning the title of the person who buys. Board level, chief executive and their direct reports are a high call point. Director or manager is a low one, and it signals that you are solving a problem an executive has already delegated, which makes the work harder to scale and the firm harder to sell. The second is client return on investment. A client who spends 500,000 dollars and can point to a five million dollar benefit gives you a clear ten times return. A client who spends the same and reports better-trained employees does not, because the benefit is real but unquantified. Segments where the return is provable are worth more than segments where it is not.
Defining an ICP is partly defining what the client would do instead. For boutiques, the competitive set is not what founders assume. Roughly forty percent of the time the competitor is doing nothing, which happens because the problem was not urgent. Around thirty percent of the time it is the client own internal team, which happens because there was no deadline. About twenty percent is other boutiques, and only about five percent is the market leaders, though those are the year-making deals.
Seventy percent of your competition is therefore inertia. That is an ICP finding, not a sales finding. If a segment keeps choosing to do nothing, it is telling you the problem is not urgent for them, and the answer is usually to change the segment rather than to push harder on the pitch.
The traditional stack ran from total addressable market to serviceable addressable market to serviceable obtainable market to an ideal client profile. It was the only viable framework at the time, but it had one flaw: it stopped at the ICP. One profile, one message, one assumption set, one ideal buyer. That is a mass-market approach in boutique packaging. Era 2 added more data and more CRM fields without changing the foundation, because humans were still building static segments from limited attributes.
Era 3 breaks that ceiling. Behavioral micro-segmentation makes it possible to find sub-segments inside your ICP, score prospects on fit and intent rather than firmographics alone, and surface fit patterns the firm did not know it had. One ICP becomes many micro-ICPs that adapt as behavior changes. This matters more for a boutique than for a large firm, because a boutique does not need thousands of clients. It needs a few of the right ones, and those are found through fit rather than volume.
What does not move is the decision itself. AI can generate segment hypotheses, pressure-test the language inside each one and show you where you are winning. It cannot decide what the firm will refuse to be. Exclusion, the choice about who you will not serve so that you can be meaningfully different to the people you will, is founder work and stays founder work.
A firm in its first year should hold the ICP loosely. You are still learning what the market will pay for, and committing early to a profile derived from three clients encodes your sample rather than the market.
Narrowing is also not free. A firm that specializes on all five dimensions at once can end up in a market too small to support it, and no one sets out to be the best of a bad bunch. Run the sizing before you commit, not after. Revenue mix constrains re-targeting too: a healthy balance is roughly sixty percent of fees from existing clients and forty percent from new, so a firm that finds a better ICP and redirects everything toward new-client acquisition trades one fragility for another.
Anchor the ICP to a problem that is pervasive, urgent and already funded, then describe the buyer twice: demographically, so you know who to approach, and psychographically, so you know what to say. Do the psychographic work through primary interviews rather than secondary research. Size the segment honestly, using modest penetration assumptions, and test reachability as carefully as size, because a market you cannot get to is not a market. Sharpen the profile with three to five forms of specialization, and prefer segments with a high call point and a provable client return. Remember that most of your competition is inertia, so a segment that keeps doing nothing is telling you the problem is not urgent for them. In Era 3 the single ICP can become many behavioral micro-segments, which suits a firm that needs a few right clients rather than many. The founder still owns the part AI cannot do, which is deciding who the firm will not serve.
In practice, most firms build only half of what they need. An ideal client profile in the useful sense has two halves: a demographic profile covering industry, size, geography, title and role, and a psychographic profile covering the buyer goals, priorities, obstacles, objections and emotional state. The demographic half tells you who to approach. The psychographic half tells you what to say when you get there. Services are bought by people, so a profile that describes only a company will not improve your win rate.
Narrow enough to be recognizably specialized on three to five dimensions, which are industry, function, segment, problem and geography. That level of specialization is what allows a boutique to charge more than a generalist. The limit is market size: it takes as much effort to pursue a small market as a large one, and boutiques typically penetrate less than ten percent of their target market. Run the sizing before committing to a narrower definition, not after.
Multiply the number of realistic clients by the revenue per engagement, then model modest penetration rather than a heroic share. The variable most often missed is reachability. A market can be large on paper and unreachable in practice, because the buyers have gatekeepers with gatekeepers. The practical fix is to publish something that lets the right buyers identify themselves, then define the target from who actually responds rather than from who theoretically qualifies.
Mostly inertia. Around forty percent of the time the competitor is the client deciding to do nothing, which happens when the problem is not urgent. Roughly thirty percent of the time it is the client own internal team, which happens when there is no deadline. About twenty percent is other boutiques and only about five percent is the market leaders, though those few are the year-making deals. If a segment keeps choosing to do nothing, that is a signal about the segment rather than about your pitch.
Sources: Greg Alexander, The Boutique: How to Start, Scale, and Sell a Professional Services Firm (Advantage, 2020), chapter 1 on testing whether a problem is pervasive, urgent and funded, the painkiller and vitamin distinction and the Capital 54 pitch, chapter 2 on demographic and psychographic profiles and the translator story, chapter 3 on the five competitors and the share of deals lost to doing nothing and to internal resources, chapter 8 on market sizing, penetration rates, reachability and the SBI early-adopter subscriber base, chapter 14 on yield and the five forms of specialization, chapter 29 on call point and client return on investment, and chapter 32 on the sixty forty balance of existing and new client fees. Greg Alexander, The AI-Native Boutique Firm (Advantage Books, January 2027), specifically The AI Marketing Manager for ICP strategy as the foundation for hyper-segmentation, the boutique need for a few right clients rather than many, and the founder ownership of exclusion and tradeoffs, and The AI Lead Generator for why the traditional market stack stopped at a single ICP and how behavioral micro-segmentation changes that in Era 3. The SBI and Capital 54 accounts are Greg own experience.
Collective 54 is the private community for founders and executives of boutique professional services firms between $5M and $50M in revenue. Members work these answers against their own numbers.