02 · The problem before the idea

The historical models of the innovation process describe two opposite directions of travel. The technology push model, associated with Vannevar Bush's report Science, the Endless Frontier (1945), describes a linear flow that starts from basic research, passes through applied research and development, and ends in the market: science produces knowledge, knowledge produces technology, technology finds uses. The market pull model, formulated in reaction to the first (Schmookler, 1966), reverses the arrow: it is demand, the need expressed by the market, that directs technological development; invention follows the economic opportunity.

Later research has shown that both linear models describe real processes badly. The chain-linked model of Kline and Rosenberg (1986) documents that innovation proceeds by iterations with continuous feedback: from the market to the design, from the design to research, from production back to the design. Scientific knowledge is not a source upstream but a reserve on which the process draws at every point when it meets an obstacle.

If the linear models are descriptively false, they nevertheless remain diagnostically useful, because the starting point of a project determines which risk dominates. A project that starts from the technology (a laboratory has developed a material, an algorithm, a process, and is looking for somewhere to apply it) typically has reduced technological risk and market risk still intact: the question "does it work?" already has a partial answer, the question "does anybody need it, enough to pay for it?" has not yet been asked. A project that starts from the problem has the opposite structure. Startup mortality leans clearly towards the first side: the recurring analyses of causes of failure (among them the series published by CB Insights on the post-mortems declared by founders) put the absence of a market need stably in first place, well above technical failures. Building what nobody wants is the most frequent and most costly error in the sector, and it is an error you commit in full before you notice it, because technology that works produces signals of progress (prototypes, demos, patents) even when it is advancing towards a market that does not exist.

Hence the methodological rule that gives the chapter its title: the problem comes before the idea. Not because ideas do not count, but because an idea is an answer, and the quality of an answer cannot be assessed until the question has been formulated precisely.

A problem, like a technological hypothesis, can be validated or refuted. There are four criteria, and each is measurable or at least estimable with evidence gathered in the field.

Frequency. How many times the problem occurs, for how many subjects. A serious but rare problem and a slight but daily problem have completely different economics; frequency multiplied by severity is a first approximation of the size of the pain, and therefore of the potential market.

Cost. How much the problem costs those who suffer it, in money, time, risk or quality. The cost must be measured on the real process, not stated in the abstract: the breakdown into stages in the previous section serves exactly to locate and quantify the cost.

Willingness to pay. The decisive question, and the most difficult, because the stated answer is systematically unreliable: almost everyone declares an interest in a hypothetical solution, far fewer buy it when it exists. The strong evidence is behavioural: is somebody already paying today for partial or improvised solutions? Do they devote their own time and resources to working around the problem? Do they sign a letter of intent, a pre-order, a paid pilot? The hierarchy of evidence runs from the stated (weak) to the contractual (strong), and chapter 07 will take up the theme within the general frame of hypothesis validation.

Reliability of the source. Who brings the problem, and from what position? A problem reported by someone who has suffered it daily for years, inside the production process in which it lives, has a different status from a problem hypothesised while reading sector reports. The best source is the direct bearer with long experience and a concrete interest in the solution; the worst is the intuition of someone who has never seen the process. Between the two there is a gradation that has to be declared and weighted.

A fifth, cross-cutting criterion concerns any solution that already exists: if the problem is real and nobody has yet solved it, the question "why?" deserves an explicit answer. Acceptable answers are of three kinds: the enabling technology is recent, the problem is recent, or the existing solutions are demonstrably ineffective or inefficient at an identifiable point. If none of the three holds, the most probable hypothesis is that the problem does not pass the first three criteria, and that the silence of the market is information, not opportunity.