Most advice about targeting assumes you already have an ideal customer profile and just need to describe it better. That's not the position most founders are actually in during the first few weeks. You have three or four plausible buyers, no evidence for any of them, and a Tuesday afternoon in which you have to decide who gets the first call. Nothing in the standard advice tells you how to make that first choice, because the standard advice is written for the step after this one.
The biggest market on paper is often the worst place to start
Left alone, most founders default to whichever segment looks largest, because bigger feels like more upside and less risk of picking wrong. It's usually the opposite bet. The largest, most obviously valuable segment is also the one with the most existing options, the most layers of approval, and the longest cycle before anyone tells you no, let alone yes. You'll spend six weeks getting a first call with an enterprise buyer who was never going to move fast regardless of what you built, and you'll come out of it with almost nothing you can use, because a slow no teaches you less than a fast one.
Segment size is a real input eventually. It's close to useless as the deciding factor for who you talk to this week, because this week you're not trying to find your biggest market. You're trying to get a fast, honest answer from somebody, and speed of answer is worth more right now than size of company.
Start where you already have an unfair advantage
The better first filter is reach, not size: who will actually take your call because of who you are, not because of what you're selling. A former colleague's company. An industry you used to work in and still understand the internal language of. A community where you're a known, credible person rather than a stranger with a cold subject line. None of that has to be your eventual market. It just has to get you a real conversation faster than a cold approach to a stranger would, because the thing you're short on at this stage isn't a theory about the market, it's actual data from actual conversations, and the fastest route to real conversations is people predisposed to give you one.
Cross that against urgency. Reach gets you the meeting. Urgency decides whether the meeting produces anything. A contact who'll take your call but has no real pain right now will be polite and useless. A contact you had to work slightly harder to reach, but who is actively annoyed by the problem you're describing, is worth more to you than the easy meeting, because they'll tell you something true.
You're not choosing your ICP. You're choosing where to learn it.
The useful reframe is that the first segment isn't a bet on the eventual market, it's a bet on where you'll learn fastest. Different criteria entirely. The eventual ICP gets defined by evidence you don't have yet: who replies, who describes their problem in words you recognize, who asks the kind of question that means they've already tried to solve this themselves. You cannot get that evidence by staring at a spreadsheet of market sizes. You get it by having twenty real conversations with a segment chosen for reachability and urgency, and then letting what you heard tell you who segment two should be.
That also means you should expect to be at least partly wrong about your first pick, and treat that as normal rather than as a sign you did the analysis badly. The first segment's job is to generate real signal, not to be correct on the first guess.
Run two or three candidates side by side instead of committing early
Because you genuinely don't know yet, it's worth testing more than one plausible segment in parallel rather than fully committing to a single guess and finding out eight weeks later that it was wrong. Rocketship makes this cheap to do properly: describe a candidate buyer in plain English, see the real count and a first sample of names at no cost, then repeat the same exercise for a second and third candidate. You'll often find one comes back as four names, one comes back as forty thousand, and one comes back as a genuinely workable few thousand, and that alone is real information about which guesses were closer to reality, gathered in an afternoon rather than a quarter.
From there, write to a small batch in each candidate group and see, honestly, which one answers. Not which one you'd prefer to be right about. Which one actually writes back, and what they say when they do.
Let the replies pick your next move, not your preferences
The hardest discipline here is following the data when it points somewhere less exciting than the segment you originally wanted to be true. Founders often have a segment in mind before they've tested anything, usually because it's the one that makes the best story, and they keep working it past the point where the replies have already answered the question. If a less glamorous segment is replying at a real rate and describing their problem in specific, recognizable language, that's your answer, whatever you'd hoped it would be.
The wrong signal to chase: whoever is easiest to reach
Reach and urgency together are the filter, not reach alone, and it's worth separating the two because founders quietly collapse them into one thing. It's tempting to work down a list of easy conversations, old colleagues, warm introductions, people who reply fast because they like you, and mistake a full calendar of pleasant calls for progress. A pleasant call with someone who doesn't actually have the problem teaches you almost nothing, no matter how easy it was to get on their calendar. If you notice you're picking who to call based on who's most comfortable to talk to rather than who's most likely to have a real, current version of the problem, that's worth catching early, because it's a way of avoiding the harder, more useful conversations without feeling like you're avoiding anything.
What this first stretch is actually for
You will not know your real ICP from a whiteboard session. You'll know it from a stack of actual replies, some encouraging and some blunt, from a segment you picked because you could reach them quickly and because they seemed to have the problem badly. Pick on those two things this week. Let the market-size conversation wait until you have evidence worth sizing.