The idea behind the curve
The theory of diffusion asks a simple question: how does something new move from a few people to many? The best-known framework comes from the sociologist Everett Rogers, whose book Diffusion of Innovations was first published in the early 1960s. Rogers drew on earlier work in rural sociology, including studies of how farmers took up new seed varieties and techniques, and generalised it to ideas, technologies and practices of many kinds.
The central picture is an adoption curve. If you plot the share of a group that has taken something up over time, the line often looks like a stretched letter S: slow at first, then steeper, then flattening as nearly everyone who is going to adopt has done so. If you instead plot the number of new adopters in each period, the same story tends to look like a bell shape. These pictures are what people usually mean by diffusion curves.
The five adopter groups
Rogers sorted people by how early they adopt, using labels that are now widely borrowed:
- Innovators are the first to try something. They tend to be adventurous and willing to accept uncertainty or failure.
- Early adopters adopt soon after, and are often respected opinion-givers whose choices others notice.
- Early majority adopters are more deliberate. They tend to follow once the idea has some evidence and social proof behind it.
- Late majority adopters are often more sceptical and may adopt because it has become the norm or because there is pressure to.
- Laggards adopt last, if at all, sometimes because of constraints, sometimes because of preference.
These are categories imposed on a continuous spread of behaviour. The usual presentation assigns each group a share of the population, but those shares come from an idealised bell curve and should not be read as measured facts for any particular trend.
What shapes the speed
Rogers and later researchers pointed to features of the innovation itself that often matter: whether it offers an advantage over what people already do, whether it fits existing habits and values, how complex it is, whether it can be tried out cheaply, and whether its results are visible to others. An idea that scores well on these tends to spread faster, although none of this guarantees an outcome.
The social side matters too. People often adopt after hearing from someone they know or trust, and communication channels, group norms and the standing of the people involved all influence how fast something travels. Mass media can spread awareness quickly, while personal contacts often matter more for the decision to act.
Where real spread differs from the model
The curve is a useful starting point, but real cases often depart from it in several ways.
Not everything reaches the end. The tidy S-curve assumes the idea eventually spreads widely. Many innovations stall early, stay niche or fade. Because stories of success are the ones retold, examples of the full curve are overrepresented, a problem related to survivorship bias.
Adopters are not fixed types. A person can be an early adopter of one thing and a late adopter of another. Labelling people rather than decisions can hide this.
Groups differ in structure. The model works best when you can define who is in the population. In practice the boundaries are unclear, and spread inside tightly knit groups can look very different from spread across a loose population. Which group you choose to measure shapes the curve you see, an issue also raised in discussions of sampling frames.
Adoption is not a single event. People try, stop, return and use at different intensities. Counting a one-time purchase or sign-up as adoption can overstate lasting use.
Curves can have gaps or reversals. Some researchers have argued that there is a gap between enthusiastic early users and the more pragmatic mainstream, so growth can stall at that point. Others find no such gap in certain cases. It is better treated as a possibility than a rule.
Network effects can bend the shape. When the value of something grows as more people use it, spread can accelerate unevenly. This is covered more fully when looking at tipping points.
Early adopters can mislead. Because the first users are often unusual in income, interests or location, their experience is not a reliable forecast for everyone else, which is the concern behind selection effects.
Using the model sensibly
Diffusion thinking is most useful as a vocabulary and a set of questions rather than a prediction engine. It prompts you to ask who has adopted so far, who has not, what might be holding back the next group, and whether the evidence of spread is about actual use or merely attention.
It is also worth checking what a claimed position on the curve is based on. Statements that a trend is “at the early majority stage” are only as good as the underlying data, such as survey results or sales records, and those depend on who was counted and how. Without a clear definition of the population and of adoption, the label is more of a story than a measurement.
Finally, remember that the curve describes spread, not worth. A fast-spreading idea is not necessarily a good one, and a slow one is not necessarily weak. Many valuable changes took a long time to move through groups, and some harmful or trivial ones moved quickly.