Ask ten B2B companies for their ideal customer profile and eight will send you something like mid market technology companies in North America. That is not a profile. It is a description of roughly forty thousand businesses, most of which will never buy anything from you, and it gives whoever is writing your outreach nothing to work with.
A profile earns its name when it excludes people. If applying it does not take a large majority of the market off the table, it has not done its job, and every campaign built on top of it will spend money reaching companies that were never going to convert.
Here is how to build one that actually narrows the field, using evidence you already have rather than a workshop and a whiteboard.
Start with the customers you already have
The pattern is usually sitting in your own account list, and it is far more reliable than anything a planning session will produce. Pull your last twenty to thirty closed deals, or as many as you have, and sort them by the things that actually matter to your business: how fast they closed, how much they pay, whether they renewed, and how much support they consume.
Then look at the top group and the bottom group separately. The question is not what do our customers have in common, because your best and worst customers usually share plenty. The question is what separates the ones you want more of from the ones you quietly regret. That difference is your profile.
If you have very few customers, use lost deals instead. Deals that reached proposal and died tell you almost as much as deals that closed, particularly when the same reason keeps appearing. No budget at that size, or the incumbent contract had two years left, are both profile criteria wearing a disguise.
Write down the firmographics that predict something
Firmographics are the structural facts about a company: headcount, revenue, industry, region, growth rate, funding stage, ownership. They matter because they are easy to filter on at scale, and because they often correlate with the thing you actually care about but cannot see directly.
Headcount is usually a proxy for whether your problem exists yet. A company of twelve does not need what you sell if it only becomes painful at eighty. Funding stage is a proxy for budget and urgency. Region determines timezone, regulation, and how a first message should read. Industry matters less than most people assume, unless your product is genuinely vertical.
Be ruthless about the boundaries. Fifty to five hundred employees is a common range and a nearly useless one, because a fifty person company and a five hundred person company buy in completely different ways. Narrow it until your outreach can speak to one buying situation rather than three.
- Headcount, stated as a tight range rather than a comfortable one
- Revenue or funding stage, where you can see it
- Region, including which timezone the buyer actually works in
- Growth signals: hiring pace, new offices, recent raises
- Ownership: founder led, private equity backed, or public
Add the criteria that are harder to filter on
The firmographic layer gets you to a long list. What turns it into a good list is the second layer: the circumstances that make your product relevant to this company right now. This is where most profiles stop short, and it is where most of the value sits.
Technology in use is the clearest example. If your product integrates with a particular platform, or replaces one, or only makes sense for companies running a certain kind of infrastructure, that is a filter worth more than headcount. So is organisational structure. The existence of a dedicated security function, a revenue operations team, or a head of partnerships tells you the company has reached the stage where your problem has an owner.
Then there are the events. A funding round, a new executive in the function you sell to, a compliance deadline landing in their sector, a wave of hiring that implies a project. None of these are permanent attributes, which is exactly why they are useful: they tell you when to reach out, not only who to reach out to.
Map the buying group, not the buyer
Almost nothing in B2B is bought by one person any more. For a typical considered purchase you are dealing with several people who have to agree, and they each need a different argument. Writing a profile that names a single job title quietly assumes a world that no longer exists.
So write down the roles. Who feels the problem day to day. Who holds the budget. Who has to technically approve it. Who can kill it without ever taking a meeting, which in most companies is security, legal, or procurement. Then decide who you open with, which is not always the most senior person on that list.
In smaller companies you go to the top, because the founder decides and everyone else implements. In larger ones, opening with a C level executive usually means being routed down two levels anyway, and you arrive at the practitioner with less credibility than if you had started there. Working upward from someone who owns the problem is often the faster route.
Define the negative profile too
The exclusion list is the least glamorous part of this exercise and one of the most valuable. Write down who you will not contact and why: companies below a size where the problem does not exist, sectors where your compliance posture cannot support them, regions you cannot service, and the customer types that have historically churned.
Then add the operational exclusions before any campaign goes live. Current customers, open opportunities, partners, investors, and anyone your team is already in conversation with. Every outbound programme eventually sends a cold email to an existing customer, and it is always avoidable and always embarrassing.
Treat the first six weeks as a test of the profile
Here is the part that gets skipped. Whatever you write down at the start is a hypothesis. It is built on a small sample of closed deals and a reasonable amount of guesswork, and the market has not yet had a chance to disagree with you.
Reply patterns are the correction. When one segment is replying at three times the rate of another, that is the profile telling you something you did not know. When a segment produces interested replies that keep failing the same qualification criterion, that segment probably does not belong in the campaign, however good it looked in the spreadsheet.
Give it enough volume to be meaningful. A few hundred contacts is not a sample, it is an anecdote. But by around six weeks of steady sending, the patterns are usually readable, and that is the moment to revise rather than the end of the quarter.
The short version
A good ICP is short. One page, written in plain language, with numbers where numbers belong and no adjectives doing work that criteria should be doing. Anyone on your team should be able to look at a company and say yes or no against it in under a minute.
If your current profile cannot pass that test, the fix is usually not more research. It is deciding to exclude more people.