05 · How an innovation spreads

Everett Rogers (Diffusion of Innovations, first edition 1962) defines diffusion as the process by which an innovation is communicated over time, through certain channels, among the members of a social system. From sixty years of empirical studies on cases ranging from hybrid seed to medicines to health protocols, two results hold with particular stability.

The first is the temporal segmentation of adopters. With respect to the moment of adoption, the population is distributed approximately along a bell curve in five categories: the innovators (about 2.5 per cent), inclined to risk and equipped with the resources to absorb failures; the early adopters (13.5 per cent), who are the decisive category because they act as the opinion leaders of the social system; the early majority (34 per cent), which adopts deliberately before the average; the late majority (34 per cent), which adopts through social pressure or necessity; the laggards (16 per cent), oriented towards tradition. The percentages are statistical conventions, not laws; what counts is the structure: the categories have different motivations, risk tolerance and sources of information, so the communication and the product that convince one category do not automatically convince the next.

The second result concerns the innovation itself: five perceived attributes explain much of the variance in the speed of adoption. Relative advantage (how far the innovation is perceived as better than what it replaces); compatibility (with existing values, experience, practices and infrastructure); complexity (negatively correlated with adoption); trialability (the possibility of trying it on a limited basis before committing); observability (the visibility of the results to third parties). The operational usefulness of the list lies in an asymmetry of cost: increasing relative advantage typically requires R&D, while improving trialability and observability requires product and go-to-market choices that are often inexpensive (trial versions, pilot projects, published case studies, referenceable installations that can be visited). Many technically superior innovations spread slowly because of defects in the cheap attributes, not in the costly one.

S-curves. The performance of a technology with respect to cumulative development effort typically follows an S-curve (Foster, 1986): slow progress at the beginning, when basic problems absorb the effort; acceleration when the architecture stabilises; saturation when the technology approaches its physical or economic limits. Technological transitions are crossings of curves: the new technology starts below the mature one and overtakes it if its curve rises faster. For timing, the consequence is that the question "is our technology better?" is badly put; the correct one is "at which point on the respective curves are we, and with what slopes?".

Dominant design. Utterback and Abernathy (1975) describe the life cycle of an industry in three phases: a fluid phase, with high product variety and competition between alternative architectures; a transitional phase, in which a dominant design emerges and becomes the de facto standard; a specific phase, in which variety collapses and competition shifts to cost and process. The moment when the dominant design emerges is a strategic watershed: entering before means risking a bet on the losing architecture; entering after means competing on efficiency against operators with an advantage of scale. The QWERTY keyboard, the architecture of the IBM PC and the VHS format are the canonical cases, and they also show that the dominant design is not necessarily the technically best one: it is the one around which the system (producers, complements, users, expectations) converges.

Network effects and lock-in. When the value of the product for each user grows with the number of users (Katz and Shapiro, 1985), directly (communications, marketplaces) or indirectly (more users attract more complements, which attract more users), competitive dynamics change in nature: critical mass, expectations ("will this standard win?") and speed count more than marginal quality; outcomes tend towards winner-take-most; and the system can lock onto solutions that are not optimal through path dependence (David, 1985; Arthur, 1989), because the coordination costs of migrating exceed the individual benefit of migrating. For a new company the preliminary diagnosis is obligatory: if the market has network effects, the diffusion strategy (a dense niche in which to reach local critical mass, subsidy of the scarce side, compatibility with the installed base) weighs as much as the product.

The industrial plan published by Tesla in 2006 ("The Secret Tesla Motors Master Plan", signed by Musk on the company site) is a rare document: a diffusion strategy declared in advance and then executed. The sequence: first build an expensive sports car in small volumes (the Roadster, 2008), use that revenue for a high-end car in medium volumes (Model S, 2012), and use that revenue for a volume car (Model 3, 2017).

Reread with the instruments of this chapter, the sequence is a manual in application. Rogers's categories: the Roadster sells to innovators and early adopters, who tolerate an immature product (limited range, non-existent charging network) and pay for novelty; skimming extracts their willingness to pay and finances the descent. The attributes: observability is designed (a sports car is circulating advertising for the thesis that "electric can be desirable"), and the charging network built in the following years works on compatibility with practices of use, the attribute on which electric was weakest. The chasm and the whole product: the passage to the early majority (Model 3) takes place only when the complete product exists, that is, car plus charging infrastructure plus support, and with the references accumulated from the high-end segments, which for a status good work like sector references in B2B. The S-curve: the plan bets on the slope of the curve of lithium-ion batteries (costs falling over documented decades), entering when the technology was sufficient for the high-end niche and letting the curve carry it to the volume segment. The case does not say that the high-to-low sequence is always right (it is a skimming choice, with the conditions seen above); it says that diffusion can be designed the way a product is designed, and that is the point of this chapter.

• E.M. Rogers, Diffusion of Innovations (5th edition, 2003): the mature synthesis of sixty years of research; the chapters on attributes and on adopter categories are the operational ones.
• G.A. Moore, Crossing the Chasm (3rd edition, 2014): the chasm, the beachhead and the whole product, with updated cases.
• J.M. Utterback, Mastering the Dynamics of Innovation (1994): the extended treatment of curves, industrial phases and dominant design.
• C. Shapiro and H.R. Varian, Information Rules (1998): still the clearest guide to strategies in markets with network effects, standards and lock-in.