Customer Segmentation for Startups: The 5-Question Scoring Model

If I had to grow a startup from zero today, I wouldn't touch the product on day one. I'd do something that sounds almost too boring to matter: pick who it's for. One group of people. Not "everyone who might like it." One.

Why start there? Because I've done it this way in every company I've helped scale over the last few years, including one that went from nothing to $2.5M a year and raised from Sequoia and a16z. Same unglamorous first move every time. And I've watched the opposite just as often: good product, real users, money in the bank, dead anyway, because they aimed at the wrong crowd and never noticed.

Customer segmentation just means splitting the market into groups that buy for different reasons, then choosing the one group to go after first. That's it. You list a few candidate groups, score each one, and commit to the winner. Everything after this, your ads, your homepage, your pricing, gets easier or harder based on this one call.

Here's exactly how to do it.

Step 1: List 3 to 6 groups you could sell to

Not 20. Not one. Three to six real candidate groups. Fewer than three and you're not really choosing. More than six and, let's be honest, you're just stalling.

And please, don't split people by the boring stuff your CRM already knows: company size, job title, age. Two companies that look identical on paper can buy for completely different reasons. Split them by why they'd buy, not by what they look like.

Step 2: Score each group on 5 questions

Give every group a 1 to 5 on each of these. Five questions, five columns. Blunt on purpose, because each one is designed to kill a weak group for a different reason.

One thing before we start: these five columns are an example, not gospel. They're the ones I reach for by default, but the right columns are whatever actually decides it for your business. If paid ads are your growth plan, you'll want a hard column for "is anyone in this niche actually making money on paid ads?" and you'll weight it heavily. If you're going sales-led, deal size and how long the sales cycle runs might matter more than reach. Someone selling to enterprises cares about procurement; someone selling to consumers cares about impulse. Swap columns in and out to fit your reality. Just keep them blunt, keep them honest, and keep them to five or so, so the board still forces a decision.

1. Can you actually reach them?

Not "how many exist." How many can you put a message in front of, this month, with the money and channels you have. A giant market you can't reach is just a nice number on a slide. Can you name a list, a group, a subreddit, an ad audience? Good. Can't find them anywhere? That's a 1.

2. How badly does it hurt?

A small pain they feel every day beats a huge pain they feel once a year, because the daily one has a budget. The best sign the pain is real: they've already built some clumsy workaround. People don't duct-tape a fix for a problem they don't have.

3. Can you help them today?

Today. Not after the three features you swear you'll ship next quarter. Score what your product does right now, for a new user, with the setup you actually have. This is where founders quietly cheat. Don't be one of them.

4. Is anyone running ads to them?

This is the question most people skip, and it's the one I'd keep if I could only keep one. If other companies have been paying to reach this crowd for years, that's proof the math works, because nobody burns money on ads to people who don't buy. An empty market usually isn't a secret goldmine. It's usually a warning.

5. Can they pay, and can they say yes?

Two things at once: do they have the money, and can the person who feels the pain actually approve the spend? Money stuck behind a committee is worth less than money someone can put on a card today.

Step 3: Average the scores, pick the winner

Add up each row, divide by five, highest average wins. That group is your ICP, your ideal customer. Here's what a finished board looks like (made-up example, real math):

Read the columns, not just the averages. The big medical groups have money and pain, but a 2 on "fit now" tells the whole story: you can't serve them yet. The salons are easy to reach and easy to help, but nobody there can pay. One low score decided each of them.

Trust your gut, just don't fly blind

Here's something the frameworks won't admit: a lot of this is gut feeling, and that's completely fine. As a founder you're forced to make calls on half the information, and your gut, after months living in this market, is often sharper than any fancy forecast. That instinct is real data. Use it. Score fast and trust your quick read on most of the board.

But there's a difference between a gut call and pulling a number out of thin air. When a score really matters and you honestly don't know, don't guess, and don't freeze either. Spend twenty minutes, not two weeks: point an AI at Reddit threads, public reports, and review sites, and ask what this group actually complains about, uses, and pays for. You just turned a blind guess into an informed one.

Then, for the handful of scores that decide the whole thing, go talk to 5 to 10 real people. The first pass of this board is a hunch. After a little digging and a few real conversations, it's a decision. Skip all of it and you'll spend six months wondering why nothing lands.

What if two groups tie?

They will, because you're scoring rough numbers on partial info. When two land within about 0.3 of each other, don't debate it for a week. Pick the one that gets you to real customers and real money fastest, where you can win a small, clear corner of the market now.

The bigger group will still be there in six months, and you'll be a much stronger company when you go after it, armed with real customers and real quotes instead of a scoring sheet. Being the obvious choice for a small group beats being option number three for a big one. Every time.

The mistakes I see over and over

• Scoring the roadmap, not the product. If you catch yourself saying "once we build X," write the 2 and move on.
• Picking the group the founder came from. Being familiar is handy. It's not a reason. Score it like any other.
• Running from competition. Rivals paying for ads is good news, not bad. It means the market has money.
• Refusing to choose. "We'll lead with A but keep B warm" is how you get a homepage that speaks to nobody. Pick one. You can revisit next quarter.
• Averaging away a 1. Any column with a 1 kills the group, no matter how high the average. A 1 is a wall, not a number.

Frequently asked questions

What is customer segmentation?

It's splitting the market into groups that buy for different reasons, then choosing one group to focus on first. For an early startup, the goal isn't a pretty chart of the whole market, it's one clear pick.

How many groups should I score?

Three to six. Fewer and you're not really comparing. More and you're stalling instead of deciding.

What's the difference between a segment and an ICP?

The segment is the comparison, the ICP is the winner. You score a few groups, and the one that comes out on top becomes your ideal customer, the one you build everything around.

How do I do this with no customers yet?

Score on what you can see from the outside: who has clumsy workarounds, who competitors pay to reach, who you can actually contact. Then talk to 5 to 10 people before you commit. No customers makes this more important, not less.

Should I just pick the biggest group?

Almost never at the start. Can you reach them, does it hurt, can they pay: those beat raw size. A huge group you can't reach or that can't pay is the most expensive kind of wrong.

What's next

You've got one group. Right now it's still just a row on a board. Next, we turn it into an actual person you understand well enough to write to, their fears, their goals, the words they use, and we figure out how much of that you're guessing.

Want someone who's run this a few times to do it with you? That's basically the first two weeks of a fractional CMO engagement.