TL;DR
A customer segment is a rule for counting people, not a description of them. Write each one as a sentence that says who exactly, in which unit you count them (companies, sites, households, people) and whether they pay or only use. Keep it to three segments at most, with no overlap and at least one that pays.
A customer segment is a counting rule. Write it as one sentence that says who exactly you count, in which unit (companies, sites, households, people), and whether they pay you or only use the product. If you cannot count it, it is not a segment yet.
A segment is a rule, not a portrait
Ask a founder who their customers are and you often hear "small businesses", "freelancers" or "anyone who works remotely". Those are not wrong. They are just impossible to count, because each one covers millions of people with almost nothing in common.
"All freelancers" counts a whole country. A segment that vague produces two problems at once:
- A swollen market. You multiply a huge headcount by your price and get a number that looks like a business. It is not one, because most of those people will never hear of you, need you or pay you.
- Customers you cannot find. There is no channel, no community and no message aimed at "freelancers". There are channels for freelance translators, or freelance developers on one platform. Vague segments leave you nowhere to go on Monday morning.
The fix is not a longer description. It is a stricter one. A segment should read like an instruction someone else could follow to produce the same count you did.
The three parts of a segment
Write each segment as one sentence with three parts:
Who exactly, counted in which unit, who pay or only use.
Who exactly. Narrow enough that two people would draw the same boundary. "Dental practices" is a start. "Independent dental practices with one to three chairs" is a segment.
Which unit. This is the part most founders skip, and it is where most errors come from. You can count companies, establishments (physical sites), households or individual people. These give very different totals for the "same" customers. The right unit is the one your price is attached to: if you charge per site, count sites; if one contract covers a whole company, count companies.
Pays or only uses. Some groups hand you money. Others use the product without paying: the patients who book through your dentist software, the guests on a booking page, the employees of a company that buys seats. Both can matter, but only the first group belongs in your revenue math. If you are not sure who pays in your setup, sort that out first.
A worked example: one customer, three counts
Take a booking and reminder tool for dental practices, priced at €50 a month per practice site. All the counts below are invented for the example; finding real ones is a separate job.
Suppose that in the country you serve there are:
- 10,000 dentists (people)
- 6,000 practice sites (establishments)
- 5,000 companies, because some groups own several sites
Counted in people.
10,000 × €50 × 12 months = €6,000,000 a year
Counted in establishments, the unit the price is attached to.
6,000 × €50 × 12 months = €3,600,000 a year
Counted in companies.
5,000 × €50 × 12 months = €3,000,000 a year
Same customers, same price, three answers. The right one here is €3.6 million, because the tool is billed per site. Counting dentists instead gives 6,000,000 ÷ 3,600,000 ≈ 1.67 times too much, a 67% overstatement that nothing in the spreadsheet will flag. Counting companies goes the other way and understates by 600,000 ÷ 3,600,000 ≈ 17%, because it misses the extra sites owned by groups.
Now add a second segment the careless way. The founder writes "dental practices in the three largest cities" next to the first one. Suppose there are 2,000 such practice sites, all of them already inside the 6,000.
6,000 + 2,000 = 8,000 sites counted, against 6,000 that exist
That is 2,000 ÷ 6,000 ≈ 33% of phantom customers, created by two segments that overlap. The clean version splits on a rule that cannot overlap, for example "solo practices (one dentist)" and "group practices (two or more dentists)". Every site falls in exactly one of them.
Last, the patients. They use the booking page every day, and there may be millions of them. They pay nothing. They go in a third segment, labelled non-paying, and they never touch the €3.6 million.
Rules that keep segments honest
A short checklist before you use your segments for anything:
- Three segments at most. More than that, and you are describing a market you cannot work alone.
- No overlap. Each customer falls in one segment only. If two segments can share a member, split them on a rule that cannot.
- At least one segment pays. If none does, you have users, not a business yet.
- One unit per segment, written down. Not "businesses (or maybe their staff)". Pick one.
- No motivations. If you want to write why they buy, stop: that is a persona, and it comes next.
That last rule feels strange, because the segment ends up looking almost too plain. That is the point. It is meant to be the poor step. Every fact, figure and story comes later and is attached to it.
Why the unit matters so much
Market sizing comes in three broad flavours: top-down from published reports, bottom-up from the number of real customers multiplied by what they pay, and estimates based on the value created when the category is new (TechCrunch, 2022).
Bottom-up is the one a solo founder can actually defend, and the segment is its first brick. Get the unit wrong and every figure stacked on it is wrong in the same direction: the market, the share you think you need, the revenue target. The error is made in the first sentence and discovered three steps later.
What this method does not cover
It does not give you the count. A clean segment tells you what to count, not how many there are. Finding a number you can trust for each segment is its own piece of work.
It does not choose the zone, either. "Independent dental practices" in one region and in a whole continent are very different counts. Decide where you will actually sell before you count anything.
And it does not tell you whether they will buy. A perfectly counted segment of people who do not want your product is still a segment of people who do not want your product.
How Banast does it
In a Banast report, you get three segments at most, each written as a sentence whose key words you click to change: who exactly is counted, in which unit (companies, establishments, households, people and a few more), and whether they pay or use without paying.
The segment step uses the model without any web search, since a segment holds no facts, only a rule. The code then checks thirteen blocking rules, among them no overlap and at least one paying segment. The mandatory unit exists because of a real bug: in an early version, a segment of private investors ended up carrying a count of 2.7 million German micro-businesses, because nothing said what was being counted.
FAQ
What is the difference between a segment and a persona?
A segment tells you who to count. A persona tells you why they buy, what they fear and how they decide. If you catch yourself writing motivations into a segment, you are writing a persona. Do the segment first, because the persona needs someone real to describe.
How many segments should a solo founder have?
Three at most, and one is often enough to start. Each segment you add is another group to count, another price to check and another channel to work. With one person and limited hours, a short list you can actually reach beats a long one you cannot.
Should I include people who use my product for free?
Yes, as a separate segment marked as non-paying. They can matter, for example as the reason paying customers stay. But they never enter your revenue math, and keeping them in their own line stops them from creeping into it.
Adrien Lebas
Founder of Banast, building it alone from Avignon. Writes about what it takes to make a living from what you build.