How Do You Define an ICP When You Only Have Twenty Customers?

How do you define an ICP when you only have twenty customers?

You stop trying to find the average customer and start finding the best ones. With twenty accounts you do not have a statistically meaningful sample, and pretending otherwise produces a profile that describes everybody and helps nobody. What you do have is enough to spot which handful went unusually well.

The mistake I see most often is a founder writing an ideal customer profile from ambition rather than evidence. It names the companies they want, not the ones that actually bought quickly and stayed.

Here is how I would build one from a small number of real accounts, and what to do when the pattern is not the one you were hoping for.

What is an ICP, and what is it not?

It is a description of the companies worth selling to, in terms you can check. Andreessen Horowitz's framework puts it plainly: great ideal customer profiles consist of specific, measurable qualities. If a line in your profile cannot be verified from the outside, it is an aspiration rather than a criterion.

It is also not a persona. The same framework notes that ideal customer profiles do not include personas, though the calls you run can identify the typical roles and titles. Keeping those separate matters, because a profile answers which company, and a persona answers which person inside it.

The practical difference shows up in your marketing. The profile decides which accounts you pursue and which content topics you own. The persona decides how a page is written. Collapse the two and you end up with pages that are vaguely about everyone at a company you never chose.

Which of your customers should you actually look at?

Not the biggest, and not the loudest. The rule of thumb in that framework is to triangulate the cohorts with the biggest and fastest revenue growth, the ones with the quickest speed to purchase, and the ones with the quickest speed to growth. Three overlapping lists, and you want the names that appear on more than one.

Speed to purchase is the criterion founders skip and the one I trust most at small numbers. A customer who decided quickly understood the value without being educated, which is the single strongest signal that you were built for them.

With twenty accounts this exercise usually surfaces four or five names. That is your working set. It feels far too small to base a strategy on, and it is still a great deal more evidence than the profile you would otherwise have written from intuition.

Which traits are worth recording?

The checkable ones. The framework points at patterns in company size, industry, geographic location, revenue, and how they use your product, and lists characteristics including company size ranges, business type, geography, industries, job titles, defined problems, company specifics, technologies used, and unique buyer behaviours.

Two of those are worth extra attention. The technologies a customer already runs are the most useful trait almost nobody records, because it is both highly predictive and easy to verify before you contact anyone. And the defined problem is the trait that keeps the profile honest, because it forces you to state why this company needs you now.

Record the traits for the four or five best accounts and for the two worst, because the contrast teaches you more than the good list alone. The trait that appears in every good account and none of the bad ones is the one to build your targeting on.

What do you do when the pattern is uncomfortable?

Believe it, at least for a quarter. The most common uncomfortable finding is that your best customers are smaller, less prestigious, or in a duller industry than the ones you have been chasing. That finding is worth money, and it is usually ignored because it is not the story anyone wanted to tell.

A related discomfort in that same framework is the distinction it draws between software companies and what it calls the Lagging 70 percent, meaning more traditional enterprises. Plenty of founders build for people like themselves and discover their real buyers are nothing like them.

I had my own version of this. Running a practice from Bengaluru for clients elsewhere in the world, my best engagements have consistently been with people who had already tried to solve the problem themselves and failed, not with the larger organisations I assumed I should be pursuing.

How narrow should the profile be?

Narrow enough that it excludes real prospects, or it is not doing anything. If your profile would not cause you to turn down an inbound enquiry, it is a description rather than a decision, and it will not change how a single hour of your week is spent.

That said, at twenty customers you should hold it loosely. Write it down, put a review date on it, and treat it as your best current hypothesis rather than a finding. The point of writing it is to make it falsifiable, not to make it permanent.

The practical test is whether it changes a decision you would otherwise have made differently. Which conference, which content topics, which enquiry you reply to first. If the answer to all three is unchanged, the profile is too wide.

How do you test it before betting on it?

By talking to the accounts that did not close. Your closed won list tells you who buys. Your closed lost list tells you whether the profile is actually predictive, and it is the half almost everyone skips because it is less pleasant.

Run a handful of short conversations with recent losses and check whether they fit your draft profile. If your best fit prospects are the ones you keep losing, you do not have an ICP problem, you have a positioning or pricing problem, and those are fixed very differently.

This is the same discipline as a win loss programme, just at a smaller scale, and it is worth setting up properly once you are past the first few. I went through how in win loss interviews that improve B2B marketing.

What changes downstream once you have one?

Your site stops trying to speak to everyone. The clearest signal that a business has a real profile is that its pages name a specific situation rather than a general benefit, and that it is willing to say who the product is not for.

It also settles arguments about hiring. A profile with a short sales cycle and a self serve motion points toward marketing. A profile with a long cycle and a named buyer inside a large organisation points toward sales. I set out that decision in whether to hire a salesperson or a marketer first.

Where it gets genuinely hard is when you have two profiles that both work. That is a real situation, not a failure, and the answer is usually separate pages rather than a blended message, which I worked through in designing one page for two different buyers.

When should you change it?

When the evidence changes, not when a quarter is disappointing. A bad month is noise. Three consecutive wins outside your profile is a signal, and so is a pattern of your best fit accounts churning after six months. Those two are worth watching deliberately rather than noticing late.

Put a date in the calendar rather than relying on noticing. Once a quarter, redo the triangulation with whatever accounts you have added, and see whether the same names and traits come out. At small numbers the profile will move, and it should.

What you should not do is widen it because pipeline is thin. That is the decision that feels like pragmatism and is almost always how a focused business becomes an unfocused one.

What should you do next?

List your customers, mark the ones with the fastest decision and the best retention, and write down five checkable traits they share. That is an hour of work, and at twenty customers it is genuinely the whole exercise rather than the first step of one.

Then write one sentence naming who this is not for, and put it on your site. That sentence does more filtering than any amount of targeting, and it is the part most businesses never publish.

Over six years and more than twenty five clients, the profile I would have written on day one and the one the evidence produced were not the same, and the gap between them was the most useful thing I learned about my own business. If you want a second pair of eyes on the pattern in your accounts, reach out and let's chat.