Every other domain in this course — value proposition, business model, growth — is downstream of one question: who exactly are you serving, and how well do you actually understand them? Get the customer and market layer wrong and you'll optimize a funnel for the wrong people, price against the wrong alternatives, or build features nobody asked to hire. This guide covers how to find, talk to, and size the market for real customers — not the imagined ones in your pitch deck.
Customer Segment & Segmentation
A customer segment is a group of people who share the same problem, context, and buying behavior tightly enough that one message and one product serve them all. Segmentation is the method for getting there — slicing a market by need, behavior, or context rather than by demographics alone, until each slice would respond to the same pitch.
Osterwalder's Business Model Generation treats Customer Segments as the first building block of any business model: an organization "must make a conscious decision about which segments to serve and which segments to ignore" before it can design anything else — value proposition, channels, or pricing all flow from that choice. Segments can be mass-market (one broad group, similar needs), niche (narrow and specialized), or segmented (same product family, slightly different needs, like a bank serving mass-affluent versus private-banking clients).
Rob Fitzpatrick's The Mom Test shows the failure mode from the inside: a founder convinced his segment was "advertisers" because "everyone advertises somehow." He talked to mom-and-pop shops, e-tailers, big brands, and agencies — all technically in-segment, all wanting different things. Every feature debate could be won by pointing to some customer who'd love it, and none could be proven wrong. His rule of thumb: if you aren't finding consistent problems and goals across your conversations, your segment isn't specific enough yet.
Micro-example: "Restaurants" isn't a segment — a food truck with two part-timers, a fine-dining GM running three shifts, and a 200-location franchise ops director all buy scheduling software completely differently. "Independent restaurants with 10–30 hourly staff and no dedicated HR person" is a segment you can build one product and one pitch for.
Watch out: A segment defined by industry or job title ("students," "advertisers," "small businesses") usually isn't a segment — it's a label covering five incompatible groups with different problems and different budgets. Fitzpatrick's fix: keep narrowing until everyone in the room would answer your interview questions the same way.
→ practice this in the customer-segments mission.
Persona
A persona is a concrete, evidence-based sketch of one representative customer — their situation, goals, and frustrations — used to keep decisions grounded in a real person rather than an abstraction.
Personas only earn their keep when they're built from evidence, not imagination. Teresa Torres's Continuous Discovery Habits is built around this discipline: teams that talk to customers every week, synthesizing what they learn as a group, keep their mental model of the customer current. Teams that build a persona once from a workshop and never touch it again drift — "each person will take away different points from the same customer interview," and the shared picture quietly rots. The antidote is treating the persona as a living document updated by continuous interviewing, not a poster on the wall.
Micro-example: Instead of "Sarah, 34, marketing manager, tech-savvy," a grounded persona reads: "Runs a 12-person restaurant with two locations. Currently schedules by group text and a paper binder. Lost $400 last week to a no-show she didn't catch in time. Checks her phone constantly during service." Every clause traces back to something a real person said or did in an interview.
Watch out: A persona built from demographics (age, job title, income) rather than behavior and context is decoration, not a tool — it tells you nothing about what to build. Personas should answer "what would make this person choose us over their current alternative," not "what does this person look like."
→ practice this in the customer-segments mission.
Jobs to Be Done
Jobs to Be Done (JTBD) frames demand as the progress a customer is trying to make in a situation — the job they'd "hire" a product to do. It's the antidote to defining your market by product category or customer demographics.
This is the entire premise of Christensen, Hall, Dillon, and Duncan's Competing Against Luck. Their famous example: a fast-food chain wanted to sell more milkshakes and tried the obvious levers — more chocolate chunks, thicker shakes, cheaper price — based on demographic data about who buys milkshakes. None of it moved sales. Then researchers asked a different question: what job did people hire a milkshake to do? It turned out half of milkshakes were bought in the morning, alone, by commuters who wanted something filling enough to make the commute less boring and neat enough to drink one-handed while driving — competing not against other milkshakes but against bananas, bagels, and Snickers bars. The other half were bought by parents in the afternoon, hired for a completely different job: being a harmless treat to make a kid feel indulged. Same product, two unrelated jobs, two different competitive sets.
The core insight: "when we buy a product, we essentially hire something to get a job done. If it does the job well... we hire that same product again. And if the product does a crummy job, we fire it." Jobs have functional, emotional, and social dimensions, and customers are willing to pay a premium for a product that nails the job precisely because the cost of a product that fails — wasted time, frustration, a bad decision — is often much higher than the price itself.
Watch out: Segmenting by demographics or product category ("people who buy milkshakes," "restaurants") will never reveal a job. You find jobs by asking what progress someone was trying to make in a specific moment, not who they are on a form.
→ practice this in the jobs-to-be-done mission.
Alternatives & Competitive Landscape
Alternatives are whatever customers use today to get the job done — competitors, spreadsheets, interns, or doing nothing — which sets the bar your product must beat. The competitive landscape is the fuller map of direct competitors, substitutes, and non-consumption you compete against for the same job and budget.
April Dunford's Obviously Awesome makes alternatives, not features, the starting point of positioning: "the features of our product and the value they provide are only unique, interesting and valuable when a customer perceives them in relation to alternatives." She tells the story of a database company that asked customers what problem they were solving and got specific, technical answers — "we need to retrieve data quickly." Useful, but not decisive. The revealing question was different: what would you use if our product didn't exist? The answers weren't other databases — they were business intelligence tools and data warehouses. That's when the team understood how customers actually categorized them.
Dunford's method: ask your best-fit customers what they'd do without you, and don't stop at "a competitor." The honest answer is often "a spreadsheet," "hire an intern," or "do nothing" — and each of those implies a different pitch. Group alternatives into clusters (e.g., "do it manually" vs. "use a small-business tool") rather than listing every possible competitor; most teams land on two to five clusters that matter.
Micro-example: A scheduling tool's true competitive landscape isn't just other scheduling apps — it's a paper binder, a group text thread, and the GM's memory. Beating "free and familiar" on price is impossible; beating it on time saved and fewer no-shows is the real fight.
Watch out: Listing every company that could theoretically compete with you (the analyst's competitive landscape) is different from listing what your best-fit customers would actually switch to (the positioning-relevant one). Only the second list should drive your roadmap.
→ practice this in the alternatives-audit mission.
Urgency & Trust
Urgency is how badly and how soon a customer needs the problem solved — the gap between "nice someday" and "budget approved this quarter." Trust is the customer's belief that you'll deliver what you promise, built through proof, credibility, and reduced risk — and it's required before money moves.
Geoffrey Moore's Crossing the Chasm explains why trust is structural, not just a feeling: high-tech markets are made of people "who reference each other when making a buying decision." Early adopters (visionaries) buy on their own intuition with no reference base — they don't want a well-tested product with a long client list, because that would mean they're too late. Pragmatists (the early majority) are the opposite: they will not buy without references from people just like them, and visionaries make terrible references for pragmatists because pragmatists don't trust their judgment. This is the mechanical root of "the chasm" — a company that nailed trust with early adopters can still stall completely with the mainstream market because it built the wrong kind of trust.
The Mom Test adds the urgency side: don't trust what people say about the future, trust what they've already spent time, money, or reputation on. Fitzpatrick's "commitment and advancement" test — will they introduce you to their boss, pre-pay, or block time on their calendar — separates a customer who's merely being polite from one who's actually urgent enough to act. "Compliments are worthless and people's approval doesn't make your business better."
Watch out: Confusing enthusiasm with urgency (customer says "I'd definitely buy that!") or confusing a reference from an early adopter with proof the mainstream market will trust you — Moore's whole chasm chapter is about founders who made exactly this mistake and stalled.
→ practice this in the urgency-and-trust mission.
Problem Interview
A problem interview is a conversation that digs into a customer's past behavior around a problem — what happened, what they tried, what it cost — without pitching your solution.
This is the subject of Fitzpatrick's The Mom Test, named for a set of rules simple enough that even your mom, who wants to be supportive, can't lie to you with her answers: talk about their life instead of your idea, ask about specifics in the past instead of generics or opinions about the future, and talk less than you listen. "Would you buy a product which did X?" fails the test — it's a hypothetical about the future, and people are "wildly optimistic about what they would do" in an imagined future. "Talk me through the last time that happened" passes — it forces a specific, checkable memory instead of a compliment.
The mechanism is subtle: by never mentioning your idea, you remove the customer's incentive to be nice to you, and you stop leading the witness. You find out whether people actually care by asking what they've already done about the problem, not by asking whether they'd like your solution.
Micro-example: Bad: "Would you use an app that made restaurant scheduling easier?" (invites a compliment). Good: "Walk me through how you built last week's schedule — what took the longest?" followed by "What have you tried before this?" (surfaces real behavior and cost).
Watch out: The most common failure is mentioning your idea too early, even in passing — the moment you do, every answer that follows is contaminated by the customer's desire to be encouraging rather than honest.
→ practice this in the problem-interview mission.
Early Adopters
Early adopters are the customers who feel the problem so acutely they'll tolerate a rough product now — your first market and your best source of learning.
Moore's technology adoption life cycle places early adopters right after the tiny group of technology-obsessed "innovators": they're not technologists, but they can "imagine, understand, and appreciate the benefits of a new technology" and act on intuition rather than waiting for references. They're also the least price-sensitive segment on the whole curve, because they see enough strategic upside to fund the risk themselves. That combination — high pain tolerance, willingness to act without proof, budget to spend — is what makes them buildable-for before you have a mature product.
Peter Thiel's Zero to One gives the sharpest illustration of choosing an early-adopter segment deliberately. PayPal didn't try to win "everyone who sends money" — they targeted eBay "PowerSellers," roughly 20,000 professional sellers with a constant stream of transactions and a terrible existing payment experience. That small, underserved, high-urgency niche became "extremely enthusiastic early adopters," and dominating that beachhead became the platform for everything after. The lesson generalizes: an early-adopter segment isn't "whoever will say yes" — it's whoever has the most acute version of the problem and the standing to act on a hunch.
Watch out: Mistaking a large, easy-to-reach audience for your early-adopter segment. The right early adopters are usually a small, specific, sometimes obscure group — not the biggest slice of your eventual market.
→ practice this in the customer-segments mission.
Willingness to Pay
Willingness to pay is the maximum a customer would actually part with for the outcome — revealed by behavior and pricing tests, not by asking hypothetically.
Competing Against Luck ties this directly back to Jobs to Be Done: customers pay a premium "because the full cost of a product that fails to do the job — wasted time, frustration, spending money on poor solutions — is significant." A product that solves the job precisely earns pricing power that a "cheaper and crappier" alternative never will, because customers are pricing in the cost of not having their job done well, not just comparing sticker prices.
The Mom Test is the warning label on how not to find this number: "How much would you pay for X?" fails the test outright — it's a hypothetical, and answers to hypotheticals are cheap to give and mean almost nothing. Fitzpatrick's fix is to look for real signals of value instead: has the customer already spent money trying to solve this problem? Would they pay a deposit right now? What's it costing them today to go without a solution? Those are facts you can build a price around; a number pulled from a survey question is not.
Micro-example: Asking "would you pay $50/month for this?" gets an enthusiastic "sure!" Asking "what have you already paid to solve this, and what did it cost you when it went wrong last month?" gets "we paid a contractor $600 to fix the mess after a double-booking" — a number you can actually price against.
Watch out: Treating a survey answer or a verbal "I'd pay for that" as validated pricing data. Willingness to pay is only real once it's backed by a commitment — money, a signature, or a deposit.
Switching Costs
Switching costs are the time, money, and pain a customer incurs to move from their current alternative to yours — friction your value must overwhelm before they'll switch.
Hamilton Helmer's 7 Powers breaks switching costs into three types, worth knowing because they compound differently. Financial switching costs are transparently monetary — new software licenses, lost sunk investment. Procedural costs come from lost familiarity and retraining — employees who've learned one system resist relearning another, independent of any feature gap. Relational costs come from emotional bonds to a vendor's people or community — a customer who likes their account rep or feels part of a user community resists leaving even when a better product exists. Helmer's HP/SAP example is concrete: when HP was forced to migrate a $7.5 billion division to SAP, the migration disruption alone cost $160 million — "considerably more than the software itself."
Switching costs cut both ways for a startup. They're the wall you have to climb to win a customer away from an incumbent (which is why "10x better, not 10% better" matters when you're the challenger). And once you've won a customer, the same three types of cost become your retention moat — which is why vendors deliberately deepen integration and build add-on products over time.
Micro-example: A restaurant that's spent six months getting staff trained on a scheduling tool, has three months of shift history logged in it, and has a good relationship with their account manager faces real financial, procedural, and relational costs to switch — even to a strictly better product.
Watch out: Underestimating procedural and relational switching costs because they're invisible on a spec sheet. The incumbent you're trying to displace is rarely losing on features alone — it's usually winning on inertia you haven't priced in.
Market Sizing & TAM / SAM / SOM
Market sizing is estimating how many customers exist and what they'd pay, built bottom-up from segment counts and prices rather than top-down from industry reports. TAM / SAM / SOM is the standard three-ring frame for presenting that estimate: total addressable market, the serviceable slice you could actually reach, and the obtainable share you can realistically win soon.
The discipline connects straight back to customer segments: you can't size a market you haven't segmented. Business Model Generation frames the Customer Segments block as the foundation the rest of the model is built on — and market sizing is just that segment definition multiplied by segment count and price. A credible SOM is customers-in-your-beachhead × realistic win rate × price, not a percentage skimmed off a headline TAM number.
The TAM/SAM/SOM discipline itself is best used as a sanity check, not a growth plan. Pitching "if we just get 1% of a $50B market" is the classic failure mode — it says nothing about whether you can actually reach or convert anyone in that market. The credible version works top-down for the market context and bottom-up for the number you're accountable to: TAM situates you in a real market, SAM narrows to what your product and channel can reach, and SOM is the only number an investor should actually interrogate.
Micro-example: TAM for restaurant scheduling software might be $2,000M globally. SAM — independent restaurants in your target geography with 10+ staff — is 20% of that, or $400M. SOM — what you could plausibly win in 3–5 years given competition — is 5% of SAM, or $20M. That $20M is 1% of the original TAM, which is exactly why investors skip the big circle and ask about the small one.
Watch out: Presenting TAM as if it were your revenue opportunity. Build SOM bottom-up from your actual beachhead segment and defend it with a customer count and a price — if you can't, it doesn't belong on the slide.
→ size your own market with the tam-sam-som calculator.
Go deeper
- Rob Fitzpatrick, The Mom Test — for interviewing technique, avoiding false positives, and reading commitment signals.
- Christensen, Hall, Dillon & Duncan, Competing Against Luck — for the full Jobs to Be Done theory, including the functional/emotional/social dimensions and the milkshake study.
- Geoffrey Moore, Crossing the Chasm — for the adoption life cycle, why references drive trust differently across segments, and the beachhead/bowling-pin strategy for expanding from an early-adopter niche.
- Teresa Torres, Continuous Discovery Habits — for building and maintaining personas and opportunity maps through continuous customer contact.
- April Dunford, Obviously Awesome — for finding true competitive alternatives and best-fit customers as the foundation of positioning.
- Osterwalder & Pigneur, Business Model Generation — for the Customer Segments building block and segment archetypes (mass market, niche, segmented, diversified, multi-sided).
- Hamilton Helmer, 7 Powers — for the anatomy of switching costs (financial, procedural, relational) as both a competitive barrier and a retention moat.