The S-Curve Does Not Care About Your Launch Date

Videoconferencing was demonstrated in 1964 at the New York World's Fair. AT&T's Picturephone. Live two-way video calls. The technology worked. The public was impressed. Then nothing happened for about four decades. Video calling did not go mainstream until the mid-2000s for early adopters and did not actually reach mass adoption until 2020, when a pandemic forced it. Fifty-six years from working demonstration to widespread use. A product manager who read the S-curve as a timeline would have been waiting a long time for the steep part.

Rogers (1962, later updated substantially through 2003) developed the diffusion of innovations framework to explain exactly this kind of puzzle. His central argument is that diffusion is a social process, not a product process. The S-curve that shows slow initial adoption, a steep rise, and then a leveling-off is not a timeline you can read off from your launch date. It is an outcome that depends on five characteristics of the innovation itself, and if any of them are wrong, the curve may never steepen.

The five characteristics Rogers identified are: relative advantage, compatibility, complexity, trialability, and observability. Relative advantage is whether the innovation is better than what it replaces, in the eyes of the potential adopter. Not in the eyes of the inventor. Not on a spec sheet. In the lived experience of the person being asked to change. Compatibility is whether the innovation fits with existing values, practices, and experiences. Complexity is how difficult the innovation is to understand and use. Trialability is whether potential adopters can experiment with the innovation on a limited basis before committing. Observability is whether the results of the innovation are visible to others.

The theories.html matrix from my study notes describes Rogers' key mechanism as: "Relative advantage, compatibility, complexity, trialability, observability, and adoption stages." The adopter categories he identified, innovators, early adopters, early majority, late majority, and laggards, describe how different types of people enter the adoption curve at different points and for different reasons. Innovators are risk-tolerant and technology-enthusiastic. Early adopters are more selective but still willing to move before the mainstream. The early majority is deliberate and waits for evidence. The late majority is skeptical and adopts mostly because of social pressure. Laggards are resistant by default and often resource-constrained.

My study notes are also careful about one trap: Rogers' categories classify individuals, not organizations. If the question is about organizational adoption, you need Cooper and Zmud (1990) or Kwon and Zmud (1987) for the staged implementation process. Rogers explains why individuals spread innovations across social systems. He is not explaining why an organization's IT department progresses through implementation phases. These are different levels of analysis, and collapsing them is a common error.

Back to videoconferencing. Run Rogers' five characteristics against the Picturephone in 1964. Relative advantage: maybe, but telephone calls were already fast and cheap. Compatibility: completely alien to existing communication practices of the time. You had to be in a special booth. The call was expensive. It required two-sided infrastructure. Complexity: very high. Trialability: almost zero. You could not try it easily. Observability: low, because almost nobody had done it, so there was nothing to observe. Four out of five characteristics were wrong. The S-curve did not steepen. That is not a market failure. That is the Rogers framework working exactly as designed.

What changed? Smartphones solved trialability. Zoom solved compatibility with existing work practices. The pandemic solved relative advantage by making any alternative worse. Three of the five characteristics shifted in a short period, and the curve steepened within months. The technology itself had not changed in any fundamental way. The five-characteristic profile of its relationship to the adopter population had changed.

I find this framework incredibly useful for diagnosing why enterprise AI tools are stalling. Relative advantage is complicated: the tools are impressive in demos but hard to connect to specific job performance outcomes. Compatibility is low: most knowledge work routines were not designed for a human-AI collaborative model, and the tooling requires people to change how they work without a clear picture of what the new workflow looks like. Complexity is high, especially in the gap between what the tools can do and what most workers know how to ask of them. Trialability is actually decent: many AI tools have free tiers and easy sign-ups. Observability is the interesting one. You can see someone use Zoom. You can tell immediately whether the video call worked. You cannot easily see whether a colleague's AI-assisted analysis was better or worse than their unaided analysis. The outcome is not visible in the same way. Low observability slows diffusion even when everything else is working.

Most product teams I have observed treat the S-curve as a vanity metric. You plot your adoption over time, look for the steepening part, and declare you are in the growth phase. What Rogers was offering is something much more useful: a checklist for diagnosing why diffusion is not happening, or why it might fail. Each of the five characteristics is a potential failure point. If your relative advantage is not real to the adopter, you have a problem. If your innovation requires people to change deeply embedded practices (compatibility), the curve will be flat. If it is too complex and trialability is limited, early adopters will not generate the word-of-mouth that pulls in the early majority.

There is a connection here to what I wrote about institutional isomorphism and AI adoption. Mimetic isomorphism can speed diffusion by making adoption feel socially necessary. When everyone in your industry is adopting a technology, the compatibility barrier is lower because the new practice is becoming the norm, not an exception. Compatible innovations, ones that fit existing norms and practices, diffuse faster through isomorphic fields. The S-curve is not purely a property of the technology. It is a property of the technology plus the social system it is being introduced into. Rogers knew this. The institutional theorists formalized it from a different direction.

I also want to separate Rogers' categories from the technology readiness literature I wrote about before. As I noted in that post, resistance is data, not an obstacle. Laggards in Rogers' framework are not the same as resisters in the technology readiness literature. Rogers' laggards are last to adopt partly because of resource constraints, partly because they observe the technology in others' hands before committing. That is actually rational behavior. Resisters in Parasuraman's TRI framework may have high discomfort or insecurity that makes them actively push back on technology regardless of its observed benefits. These are different psychological and sociological profiles, and conflating them leads to bad interventions. Telling a laggard to adopt faster does not work. Telling a resister their resistance is irrational definitely does not work.

The deeper point Rogers was making, one that I think gets lost when people reduce diffusion theory to the S-curve diagram, is that timing is not the variable you control. You cannot decide to be in the steep part of the S-curve. What you can influence are the five characteristics. You can improve relative advantage by making the benefit more concrete and demonstrable. You can improve compatibility by designing the innovation to fit existing practices rather than requiring people to abandon them. You can reduce complexity. You can create trialability. You can make outcomes more observable. These are design decisions. They determine whether the curve ever steepens. The launch date tells you when you started. The five characteristics tell you whether the clock is actually running.

The Picturephone launched in 1964. The curve started running in 2020. Fifty-six years is a long time to wait for the right compatibility frame.