13. Failure modes of business design

§The question this chapter answers: where does business design go wrong, and what did each failure look like the week before it was obvious?

§CB Insights' post-mortems put running out of money and having no business model in the top reasons startups die, right behind building something nobody wanted. The first two are the same failure seen at different distances. Five specific forms of it, in the order they tend to appear.

§ 13.1A cost invented for a problem that did not exist#

§What it looks like. The model includes a cost, a giveaway, an incentive, a subsidy, designed to make customers do something they were going to do anyway. Why it happens. The team assumes people need to be paid for what they give freely, because the team would need to be paid for it. The tell. The gave-freely pile from chapter 3 contains something the model pays for. The defense. The three piles, read before the unit economics. In the case, a point economy built to compensate guests for data that six of six handed over gladly. Removing it did not just save the cost; it removed the false assumption the tier economics rested on.

§ 13.2Pricing decided at launch#

§What it looks like. The product is built, the launch is scheduled, a price is chosen from a competitor's page the week before. Why it happens. Pricing feels like a marketing decision and marketing feels like a launch task. Ramanujam and Tacke's data says this is the single most common commercial failure. The tell. No three-number pricing conversation in any synthesis page. The pricing page is written after the product. The defense. Monetization as hypotheses, chapter 9, with the conversation and the smallest market test dated before the build.

§ 13.3A segment chosen for size#

§What it looks like. The first customer is the largest addressable group, and the first version is built for its average. Why it happens. Size feels like safety and reads well in a deck. The tell. The team cannot name a person in the segment who has shown a commitment and said a price. The defense. Three circles, chapter 10. If the overlap is empty for act one, say so and name what act one builds toward.

§ 13.4Feasibility assumed because the demo worked#

§What it looks like. The prototype tapped through beautifully, so the product is believed to be a month away and the operation is believed to be ready. Why it happens. In 2026 a working demo is the cheapest artifact in the room, and it looks like the most expensive. The tell. No technical path recommendation with cost per user. No operational number for what the floor absorbs. Staff tested in a conference room. The defense. Chapters 4 and 5. Feasible means buildable by these people with this money, and deliverable by an operation that has already shown it can.

§ 13.5Viability postponed until the product exists#

§What it looks like. Desirability and feasibility are worked hard; the business model is a slide with logos on it; the plan is to figure out monetization after traction. Why it happens. The other two lenses are more fun, and the third one might kill the idea. The tell. No unit economics sheet, or one built from category benchmarks and never opened. Paul Graham's question, default alive or default dead, has not been asked. The defense. Part III before the build. A unit that loses money is found on paper, where it costs a spreadsheet, instead of in production, where it costs the runway.

§ 13.6The root#

§Every failure above is a hypothesis that never got a number. The invented cost was an unstated assumption about what people need to be paid. The launch price was an untested assumption about value. The big segment was an assumption about fit disguised as a fact about size. The demo was an assumption about building and delivering. The postponed model was an assumption that the unit would work out.

§ 13.7What comes out#

§The five, pinned next to the list from chapter 12. Read before business design starts and again when the first pilot number comes in. Every team is in at least one. The list exists so that being in one costs a row, not a quarter.