Picking the Wrong Segment Doesn't Feel Like a Mistake
I lost about a million rubles on a pizza franchise. It took three months. The pizza was good, the franchise was solid, the ovens worked. The location was wrong. The street outside filled with retirees on fixed incomes and almost nobody who would walk in and buy a fresh pie at full price. I had picked a segment that passed my door every day and never came in, and I didn't see it until the money was gone.
I've since done about forty segmentations for large businesses, and I've watched the same thing happen to teams with far more runway than my pizzeria had. The wrong segment rarely announces itself. It pays a little, the metrics move a little, the team stays busy — and then a year is gone and nothing compounded.
The most expensive single error in product is choosing the wrong Job of the wrong segment. Every step downstream — value, pricing, communication, acquisition, retention — silently inherits it, and no amount of execution fixes a choice made wrong at the root. What follows are the five ways teams make that error without noticing, in the order I see them most.
A Paying Segment Feels Like a Solved Segment
The belief underneath all five is the same. If we picked a segment and it has paying customers, we're fine — segmentation mistakes are an academic worry, not an operational one. A segment that pays feels like a segment that's solved. It usually isn't, and the distance between pays and solved is exactly where a year of work disappears.
The Wrong Segment Always Pays Just Enough to Hide
Each of the five mistakes shares one property: it is invisible from inside while it's happening. The revenue keeps coming, the dashboard keeps moving, and nothing throws an alarm. That's what makes them quiet, and that's what makes them expensive.
The Segment That Only Lives in the Brief
The first way is to build for people who never existed — not as humans, but as buyers. A team runs interviews, hears people describe what they want, writes the Jobs down in clean language, and designs with confidence. It skips the one signal that decides everything: did these people ever pay for this in the past, with money, time, or attention? "I want to learn Spanish." "I want to lose thirty pounds." "I want to apply to business school." Each is a real wish, and most of the people saying it have never opened the app, booked the trainer, or filled out the form.
A Job nobody has ever paid for is a brief with a market-size number attached. The product converts near zero and the team can't see why, because the interviews were warm and the market study sized plausibly. This is the graveyard most aspirational products are buried in: the fitness app, the language app, the someday-MBA course, all built on I'll start Monday. The tell is easy to check and easy to skip. Ask what the person did about this last year, not what they plan to do next year.
This is a different failure from the fake criteria two chapters back. There the label named a slice and decided nothing. Here the Job itself was never performed. The fix is the cheapest discipline in research: talk only to people who already spent something on this, and if the Job they name is brand-new, climb a level, because the real one usually sits above it.
Serving Everyone Serves No One Enough to Switch
The second way is quieter: the team refuses to pick. Every customer type looks important, so the roadmap grows a branch for each. Locally every feature is justified, because someone asked for it. Globally the product becomes the seventh-best option for seven different people and the first choice for none. A product that serves everyone gives no one a reason strong enough to switch. The message blurs. Acquisition cost climbs because nobody can say in one sentence who it's for. Churn stays high because each group is only half-served, and a focused competitor sits one click away.
A brokerage I worked with in my home market — a Compass-or-Keller-Williams-shaped franchise — "served" five price classes at once, from economy to luxury. The proof came out of their own data: a single buyer showed up across all five tiers. Property class turned out to be a label on a slide, not a real boundary between customers. The operational version of the same mistake is a roadmap with branches for seven customer types and not one sentence naming the customer it's actually for.
Focus is subtraction at the level of the segment. Choosing one means removing the others from the build until the chosen one is won, not stacking it on top of everything you already do. Name the target segment out loud, cut the rest from the roadmap, and measure activation and churn per segment instead of in one comforting average.
A Segment Can Pay and Still Never Close
The third way is the hardest to accept, because the customers are real and the money is real. The segment exists, the people have the Job, they pay — and the unit economics can never work. Revenue is positive at the company level, so nothing looks broken. The cost to serve hides inside support, onboarding, custom work, and edge cases, and it surfaces only when someone computes margin per customer instead of in total.
Some segments pay and still can't be won, because the margin never closes no matter how good the product gets. The limit is set by the segment's situation: the check is too small, the support load too heavy, the frequency too low, the room to expand absent. One of the strongest founders I ever taught built a company making industrial 3D printers. The machines were excellent, and they took about half their market. Then he ran the numbers. Half the market was roughly 500 million rubles a year, with nowhere left to grow into. He closed a company that worked, because the segment was a ceiling. The same shape shows up smaller every day: the B2B tool whose customers each need a week of hand-holding to stay on a forty-dollar plan.
The discipline is to compute margin per unit, per segment, before you commit — not two years in, after the cost-to-serve has already set. Size and reachability are filters laid on top of value, not details to sort out later.
Polishing the Room You're In While a Bigger One Stays Dark
The fourth way wears the disguise of good work. The team improves the product for the customers it already has, and every number cooperates. Conversion rises, churn falls, the cohort gets happier. It feels like progress because it is — locally. What it misses is the rest of the map. An adjacent segment that's larger, pays more, or already wants Core Jobs the product performs sits unexamined, because all the effort points inward. A year spent improving the room you're in is a year not spent finding the bigger one next door.
Nobody decides to stay local. It happens by default, because tuning what exists is always the nearer task. I've watched a team spend two quarters polishing onboarding for its paying users while a segment three times larger sat one research sprint away, unlooked-at. The onboarding work was honest and the metrics moved. The bigger bet never reached the table, because no one had drawn the full map.
The last chapter sliced the segment you already chose. This mistake is about the segment you never looked at. Draw the whole map of segments around the Big Job before you pour a year into the one you're standing in. Compare them on the three things that decide it: size, margin, and room to grow.
Real Revenue That Quietly Eats Your Margin
The fifth way is the twin of the third, pointed the other direction. The third mistake is walking into a segment that can't pay off. The fifth is refusing to walk out of one you're already in. Once a product has a paying base, those customers sort themselves by how profitable and how satisfied they are. Rank them on those two axes and they fall into four groups. Call them A and B, the profitable and content, and C and D, the ones who drain support and never close the loop.
Revenue from C and D is real, and still not worth having when the cost to serve it eats the margin from the A and B customers you actually want. In most sales organizations I've audited, C and D swallow something like eighty percent of the team's time and return about twenty percent of the revenue. Teams keep them anyway, because the revenue is visible in the report and saying no feels like rejecting people you know by name. The trap is that C and D won't graduate to A and B on this product. The mismatch is in what the product was built to do, not in how well it does it. Build them a better experience and they stay exactly where they are.
Firing them is deliberate work: turn them away at the top of the funnel, automate them down to near-zero cost, or raise the price until the margin closes. One caution keeps mistakes four and five from colliding. Dropping C and D tightens the segment you already have, which is a local move. Do it — and still keep looking outward, not only in.
Wes Asks If These Are Mistakes or Just the Cost of Trying
The Algorithm Is the Answer to All Five
If these are the five ways to choose wrong, the next chapter is the sequence that chooses right on purpose.