Ideal Customer Profile

Early companies don't screen customers. Anyone willing to pay looks like the right answer, and for the first year that judgment usually holds up. The trouble starts because the customers collected that way have nothing in common with each other.

A five-person studio wants a product with nothing to configure. A three-thousand-person company wants permissions it can slice by department. Those two requests trade against each other — satisfying one makes the other worse. A team unwilling to drop either ends up carrying two sets of settings and serving both groups badly.

Sales runs into the same shape. Chase every inquiry with equal effort and the week spreads evenly across all of them, which leaves too little time for the accounts that would have stayed. Pagedeck learned this in a weekly review, when Sofia said twelve new accounts had come in that month and seven of them matched the conditions of accounts that had cancelled the previous quarter. Until the company wrote down which customers actually fit, it kept repeating the same month.

An ideal customer profile (ICP) is that description, pinned down as a document. It isn't a list of companies a team would like to sell to. It's a record of what already-successful customers had in common.

Split the existing customer list into the accounts that worked and the accounts that didn't, find what separates the two groups, and write those differences down as conditions. The method rests on evidence that already exists rather than on a guess about the market. It takes five steps.

1. Pull out the customers that worked. They have been paying for more than a year, their usage hasn't fallen off, and some have moved up a plan. Contract size is not the sort order — a large deal that ends after six months is common enough.
2. Pull out the opposite group just as deliberately. Accounts that churned within six months — that is, paid for a while and then left — plus accounts that still pay while nobody logs in. Without this group there is no line to sort against.
3. Put the two lists side by side and look for what differed. Company size, industry, tools already in use, the job title of whoever started the rollout, how they pay.
4. Keep only the conditions that resolve to yes or no. "Companies that write a lot of documents" can't be judged from an inbound message, so it never becomes a condition. "Companies already running an internal wiki" can be checked, so it does.
5. Write the exclusions separately. This is usually the half that saves the most hesitation. Pagedeck put "requires an on-premises install" on the exclusion list, and that one line ended three or four long internal debates a month.

One rule governs how the conditions get written: the subject is a company, not a person. Between 30 and 200 employees, builds software, already uses at least one collaboration tool. Who signs the contract and who opens the product every morning is the job of a buyer persona instead.

Conditions aren't written once. As the product changes and the customer list gets longer, the line separating the two groups moves, so teams run the same exercise again roughly every six months.

The most frequent use is the moment an inquiry arrives. When sales runs qualification — the process of checking whether a prospect can actually buy — this list of conditions is the first thing the inquiry gets held against. A match books a meeting. A mismatch moves down the queue or gets closed with the reason recorded.

In marketing it decides where money goes. Which industries the ads target, which conferences are worth a booth, who the writing is aimed at — all of it comes out of the same conditions.

On the product side it settles arguments. When two customers ask for opposite things, the first question is which of them fits the profile. That's what keeps the roadmap responding to the customers who will stay rather than to the customers who complain loudest, and it's why the product roadmap and the profile tend to get reviewed in the same meeting.

When several groups fit the conditions at once, there's a further step. Splitting effort across all of them wins nothing anywhere, so a team picks one segment to take first — the beachhead market.

"This one is outside our ICP. Do we still take the meeting?"

Asking whether a mismatch can wait behind the inquiries that fit.

"We moved the ICP above 30 employees and six-month cancellations dropped by half."

Narrowing the conditions raised the share of customers who stayed.

"Is the account that asked for that feature inside the ICP?"

Checking how much weight the request carries before deciding whether to act on it.

Pagedeck first wrote its conditions down right after the Series A. Mia pulled two years of accounts into two columns — kept past a year, gone within six months — and the kept column was mostly companies above 30 employees where a department head had started the rollout. The gone column was mostly single users trying the product alone, ending when that person left the job.

So "adopted at the department level" went into the conditions alongside the size range. Ravi narrowed advertising to software companies, and Sofia started asking about the rollout on first contact. Three months later the number of new contracts was actually lower, while the share still active at six months had gone up.

The most common mistake is reading the profile as a wish list of companies to sell to. The word "ideal" invites it. But the document comes from what has already worked, not from what a team hopes for next. Fill it with hope and it turns into a row of enterprise logos, and a sales plan built on those conditions reaches the end of the quarter without closing anything.

Reading it as a ban on everyone outside the conditions is the other common error. What the profile sorts is not who gets served but where limited time and money go first. Inquiries from outside the profile still close when the buyer is ready. They simply don't get chased first, and their feature requests don't get the front of the roadmap. Teams that miss this distinction start treating the document as something that blocks sales, and a few months later nobody opens it.