Diffusion of innovation describes how a new idea, product, or practice spreads through a population over time. The concept, most associated with sociologist Everett Rogers, maps out why some innovations take off quickly while others stall, and why certain people adopt early while their neighbors hold out for years. It’s one of the most widely applied frameworks in social science, used to explain everything from the adoption of hybrid seed corn in 1930s Iowa to the global spread of smartphones. But the framework also has blind spots and biases that are worth understanding, especially if you’re trying to apply it to real decisions.
The Five Stages of Adopting Something New
Rogers described adoption not as a single moment of decision but as a process that unfolds in five stages. First comes knowledge: you hear about the innovation and start to understand what it does. Then persuasion: you form an opinion about it, favorable or not. Third is the decision itself, where you commit to adopting or rejecting it. Fourth is implementation, when you actually start using it. And fifth is confirmation, where you look for reassurance that you made the right call and may reverse course if you encounter conflicting information.1TheoryHub Book. Diffusion of Innovations: A review – Section: Theory
This sequence matters because it reveals how fragile adoption can be. Someone might know about an innovation and even think favorably of it, yet never reach the decision stage. Or they might adopt, use it briefly, and then abandon it during the confirmation phase. Anyone who has downloaded an app, used it for a week, and then deleted it has lived through all five stages. The model also implies that interventions need to target the right stage: advertising builds knowledge, but it won’t help someone stuck between decision and implementation because they lack access or resources.
Who Adopts First and Who Waits
Rogers divided any population into five adopter categories based on how quickly people embrace something new. Public health literature typically cites approximate percentages: innovators at about 2.5%, early adopters at 13.5%, the early majority at 34%, the late majority at 34%, and laggards at 16%.2PubMed. Diffusion of innovations: a guiding framework for public health These numbers describe a bell curve: most people sit in the middle, and the tails are thin.
In practice, the distribution can shift depending on context. A study of smallholder farmers in Ghana found that innovators made up only about 1.4% of respondents, early adopters about 11%, the early majority about 33%, the late majority about 37%, and laggards about 18%. The early and late majority together accounted for more than 60% of all respondents, confirming that most adoption decisions in that population were made by people who waited to see what others did first.3Heliyon. Innovation adopter categories of smallholder farmers in Ghana and their innovativeness – Section: Results and discussions The researchers noted that late-majority farmers were skeptical about new technologies and wanted to see evidence of others using them successfully before committing.
The categories aren’t personality types etched in stone. Someone who adopts a new farming technique late might adopt a new mobile payment system early, depending on their access, resources, and what’s at stake. Still, the general pattern holds across domains: a small vanguard moves first, the bulk of the population follows once adoption feels safe, and a trailing group resists or delays until external pressure forces the issue.
What Makes an Innovation Spread Faster
Not every innovation is created equal when it comes to spreading. Rogers identified five attributes that shape how quickly people adopt something new: relative advantage, compatibility, complexity, trialability, and observability.4Elsevier. New approach to Rogers’ innovation characteristics and comparative implementation study – Section: Introduction Each is somewhat related to the others, but they point to different levers.
- Relative advantage: How much better is this than what it replaces? An innovation doesn’t have to be objectively better in every dimension, just clearly better in the dimensions people care about.
- Compatibility: Does it fit with existing values, past experiences, and current needs? A tool that requires people to change their entire workflow will spread more slowly than one that slots into existing routines.
- Complexity: Is it easy to understand and use? The harder something is to grasp, the slower it spreads. This is where design and communication matter enormously.
- Trialability: Can people test it without fully committing? Free trials, demo versions, and sample sizes all reduce the risk of adoption.
- Observability: Can others see the results? An innovation whose benefits are visible to neighbors spreads faster than one whose benefits are private or invisible.
These attributes explain a lot of real-world adoption patterns. Solar panels spread partly because they’re highly observable: your neighbors can literally see them on your roof. A new medication, by contrast, offers invisible benefits and often spreads more slowly among patients, even when the clinical evidence is strong. The innovation itself is not always the bottleneck; how people perceive it is what drives or stalls adoption.
The S-Curve and the Tipping Point
When you plot cumulative adoption over time, the result typically looks like an S-shaped curve. Adoption starts slowly as innovators and early adopters take the plunge, accelerates sharply as the majority joins in, then levels off as only the most resistant holdouts remain. This S-curve shows up with remarkable consistency across different types of innovations and markets.
The steepest part of the curve is sometimes called the tipping point, where adoption shifts from a niche activity to mainstream behavior. An analysis of 94 digital platforms found that their adoption patterns were typically S-shaped and that calculating the tipping point was possible, with at least 13 months of user growth needed before a platform reached it.5PACIS 2020 Proceedings. S-Curves in Platform-based Business – Facing the Challenge of the Tipping Point – Section: Abstract Before that threshold, the platform was still fragile and could easily collapse; after it, growth became self-sustaining.
The Bass diffusion model, one of the most widely used mathematical treatments of this process, separates two forces driving adoption: external influence (like advertising) and internal influence (like word of mouth from existing adopters). The interplay of these two forces generates the S-curve. Researchers have extended this model to account for successive generations of technology, where a newer product doesn’t just diffuse on its own but also substitutes for an older one.6Management Science. A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products – Section: Abstract Think of the shift from DVDs to streaming: the newer technology’s adoption curve overlaps with the older one’s decline.
How Social Networks Shape What Spreads
Diffusion doesn’t happen in a vacuum. It flows through social networks, and the structure of those networks matters as much as the innovation itself. Research consistently shows that diffusion outcomes are shaped by network properties like density, centrality, and the strength of ties between people.7Sociology Compass. Diffusion of innovations through social networks: Determinants and implications – Section: Abstract
Opinion leaders play an outsized role. These are people who hold influence within their networks, not necessarily because of formal authority but because others trust their judgment. Simulation studies have demonstrated that when opinion leaders initiate the diffusion process, adoption speeds up and the total share of people who eventually adopt increases.8Journal of Product Innovation Management. Opinion Leaders’ Role in Innovation Diffusion: A Simulation Study – Section: Abstract Researchers have even developed methods to identify these individuals and strategically use them to accelerate diffusion.9The ANNALS of the American Academy of Political and Social Science. Accelerating the Diffusion of Innovations Using Opinion Leaders – Section: Abstract
The type of social tie matters too. Strong ties, the close relationships you have with friends and family, are better at transmitting information within a group. Weak ties, your loose acquaintances, are more effective at carrying information across group boundaries because they bridge otherwise separate clusters.10Social Networks. Information flow through strong and weak ties in intraorganizational social networks – Section: Abstract The highest likelihood of information flow comes from a combination of both.
There’s also the question of whether people spread innovations more effectively to people who are similar to them (homophily) or to people who are different (heterophily). Research on word-of-mouth communication suggests the answer is somewhere in between: there’s an optimal level of difference between sender and receiver that maximizes diffusion.11Proceedings of the International AAAI Conference on Web and Social Media. Optimal Heterophily for Word-of-Mouth Diffusion – Section: Abstract Too similar, and you’re preaching to the choir. Too different, and the message doesn’t resonate. A study of solar water heater adoption in Chinese cities found that strong, homophilous ties in some communities facilitated both information flow and social conformity pressures that encouraged adoption.12Energy Research & Social Science. Social ties, homophily and heterophily in urban sustainability transitions: User practices and solar water heater diffusion in China – Section: Conclusions
Simple Versus Complex Contagion
Early diffusion models often treated adoption like a virus: if you’re exposed once, you have some probability of “catching” it. This is called simple contagion. But a growing body of research shows that many real-world adoption decisions work differently, requiring reinforcement from multiple sources before someone acts. This is complex contagion.
A country-scale field experiment tested this directly. People were randomly exposed to either one or two friends who were encouraged to share a coupon for a mobile data product. The results provided strong support for complex contagion: signals from multiple sources amplified each other through what the researchers called synergistic interdependence, meaning the effect of two friends sharing was more than double the effect of one.13Sociological Science. Complex Contagion in Social Networks: Causal Evidence from a Country-Scale Field Experiment – Section: Abstract A separate experiment on Twitter found that a complex contagion model, where adoption depends on the number of distinct sources a person hears from, fit observed information-spreading behavior more accurately than a simple contagion model.14PLoS ONE. Evidence of complex contagion of information in social media: An experiment using Twitter bots – Section: Results
Researchers have also proposed models where both simple and complex contagion dynamics can emerge simultaneously within the same social system, depending on the beliefs and cognitive mechanisms involved.15PubMed Central. Emergence of simple and complex contagion dynamics from weighted belief networks – Section: Abstract The practical takeaway is that for many innovations, a single exposure isn’t enough. You’re more likely to try something new when multiple people in your network have already done so, which is one reason why the S-curve has that slow initial phase before adoption accelerates.
Why Culture Changes the Speed of Adoption
Diffusion doesn’t happen at the same rate everywhere, and national culture is one reason why. A study of firms across Europe found that cultural variables significantly predicted differences in adoption decisions, even within the EU, where countries share common market structures and regulatory frameworks.16European Management Journal. The Influence of National Culture on the Adoption Status of Innovations: An Empirical Study of Firms Across Europe – Section: Abstract Cultural differences between countries were large enough to affect adoption above and beyond the usual firm-level and industry-level factors.
A broader cross-national study examined how cultural dimensions relate to national rates of innovation in 33 countries. The strongest predictor was uncertainty acceptance: countries whose citizens were more comfortable with ambiguity and risk innovated at higher rates. Low power distance (less rigid hierarchies) and individualism were also associated with faster innovation, suggesting that cultures encouraging people to think independently and challenge authority tend to generate and adopt new ideas more readily.17Journal of Business Venturing. Cultural influences on national rates of innovation – Section: Abstract
This has practical implications for anyone rolling out an innovation internationally. A launch strategy that works in a culture with high uncertainty acceptance and weak hierarchies may fall flat in a culture where risk aversion is high and decisions flow through gatekeepers. Compatibility, one of Rogers’ five attributes, extends beyond the innovation’s technical features to the cultural soil it lands in.
When Organizations Adopt, It’s a Different Process
Most of the classic diffusion framework was built around individual decisions: a farmer choosing to plant a new seed, a consumer buying a new gadget. But organizations adopt innovations differently. In many workplaces, employees don’t choose the new software or process themselves; someone higher up makes the decision and rolls it out. This authority-driven adoption process doesn’t map neatly onto the stages Rogers described, because the person implementing the innovation may never have gone through a persuasion or decision stage at all.18ACM SIGMIS Database: the DATABASE for Advances in Information Systems. Organizational adoption and assimilation of complex technological innovations – Section: Abstract
This gap matters in healthcare, education, and corporate settings where adoption is mandated from above. A study of e-health adoption in Mauritius found a significant association between adopter category and willingness to recommend e-health, with physicians and nursing managers playing central roles in facilitating diffusion within the healthcare system.19PubMed Central. Assessing E-Health adoption readiness using diffusion of innovation theory and the role mediated by each adopter’s category in a Mauritian context – Section: Results Even in top-down adoption scenarios, certain people serve as internal opinion leaders who shape whether the mandate succeeds or fails in practice.
Research on global health innovation adoption in the United States identified six factors that mattered for bringing internationally developed innovations into the domestic context: the innovation’s attributes, the presence of linking agents who bridge communities, inter-organizational partnerships, strategies for scaling up, implementation processes, and the broader policy and social context.20PubMed Central. A Framework for Introducing Global Health Innovations to the US – Section: FINDINGS One surprising finding was the role of stigma regarding the origin of innovations: interventions developed in lower-income countries sometimes faced resistance from U.S. adopters who questioned their quality, a factor that doesn’t feature prominently in the classic diffusion literature.
The Pro-Innovation Bias
One of the most persistent criticisms of diffusion research is that it suffers from a pro-innovation bias: the assumption that innovation is inherently good and that adoption is the desirable outcome. Rogers himself acknowledged this. The bias shows up not just in how researchers frame their work but in what they choose to study in the first place.
A study of interstate policy diffusion in the United States tested this directly. Researchers found that limiting analysis to widely adopted policies, the “successes,” skewed results in systematic ways. It led scholars to overestimate the influence of geographic proximity and policy attributes while underestimating the role of professional associations and the opportunity to learn from earlier adoptions.21Political Research Quarterly. Policy Diffusion and the Pro-innovation Bias – Section: Abstract In other words, by only studying innovations that spread widely, researchers were getting a warped picture of what actually drives diffusion.
Public health scholars have also raised this concern. The diffusion framework tends to frame non-adopters as “laggards” who are behind the curve, which can shade into individual blame. But someone who declines a new technology may be making a perfectly rational decision based on their resources, context, or past experience with failed innovations. The framework doesn’t always account well for societal, cultural, or structural barriers that have nothing to do with individual attitudes.22PubMed. Diffusion of innovations: a guiding framework for public health
Researchers have proposed supplementing the diffusion model with social network theory to fill these gaps, using constructs like reciprocity, transitivity, and network density to understand adoption patterns that individual-level frameworks miss.
Online Platforms and Network Effects
Digital platforms introduced a dynamic that Rogers’ original framework didn’t contemplate: network effects. The value of joining a social network or marketplace increases as more people join it, which fundamentally alters the adoption curve. Research on an online social network found two powerful mechanisms. First, a person connected to many existing adopters was more likely to adopt. Second, when those existing adopters were densely connected to each other (a clustering effect), the influence on potential new adopters was even stronger.23Journal of Marketing Research. Network Effects and Personal Influences: The Diffusion of an Online Social Network – Section: Abstract
This clustering effect connects to the complex contagion research discussed earlier. On a platform where your close friend group is already active, you feel social pressure and see immediate value in joining. On a platform where only one distant acquaintance is active, the pull is weak. This is why social media platforms often grow in geographic or social clusters rather than evenly across a population, and why a platform can be wildly popular in one city while barely registering in another.
When the Same Mechanisms Spread Misinformation
The diffusion framework was designed to explain how useful innovations spread, but the same mechanisms apply to the spread of misinformation. Researchers have applied Rogers’ model to examine how false claims move through digital environments, passing through the same adopter categories and following the same S-curve dynamics as legitimate innovations. Misinformation tends to spread especially well when it aligns with what people already believe, triggers strong emotions, or gets amplified by platform algorithms.24AI & SOCIETY. Cascading falsehoods: mapping the diffusion of misinformation in algorithmic environments – Section: Abstract
This isn’t just an academic curiosity. It means that the same features that make an innovation spread quickly, like observability and compatibility with existing beliefs, also make false information sticky. Algorithmic recommendation systems add a layer that Rogers never accounted for: they can artificially inflate the “social proof” around a piece of content, making it appear that many people already endorse it, which in turn triggers the complex contagion dynamics that drive adoption. The diffusion framework, applied honestly, is value-neutral about what’s spreading.
Incumbent Resistance to New Ideas
Diffusion research tends to focus on adoption, but resistance is equally important and far less studied. Established players in any field have reasons to slow or block the spread of innovations that threaten their position. Research on institutional resistance has developed a typology of how incumbents fight back, distinguishing between loud forms of resistance like public attacks and smear campaigns, and quiet forms like bureaucratic delay and strategic silence.25Organization Theory. Slander, Shouts, and Silence: Incumbent Resistance to Disruptive Logics – Section: Abstract
The quiet forms are arguably more effective precisely because they’re harder to detect. An industry group that delays regulatory approval, a professional association that declines to endorse a new practice, or a company that makes it difficult to switch to a competitor’s product: these are all forms of resistance that slow diffusion without ever appearing on a classic adoption curve. Understanding diffusion without accounting for active resistance gives an incomplete picture, which is part of why the pro-innovation bias discussed earlier persists so stubbornly in the literature.
Diffusion Beyond Humans
One of the more surprising applications of diffusion research extends to non-human animals. Social diffusion experiments have documented cultural transmission in a range of species, with innovations spreading across groups of up to 24 individuals and along chains of up to 14 transmission events.26PubMed Central. Review. Establishing an experimental science of culture: animal social diffusion experiments – Section: Abstract
A striking example comes from humpback whales. Researchers tracked the spread of a foraging technique called lobtail feeding, in which a whale slaps the water surface with its tail before lunging to feed, through a population over 27 years. Using network-based diffusion analysis, they found that support for models including social transmission was 6 to 23 orders of magnitude greater than for models without it.27PubMed. Network-based diffusion analysis reveals cultural transmission of lobtail feeding in humpback whales The technique didn’t arise independently in scattered individuals; it spread through social connections, with whales learning from whales they associated with. The diffusion curve for lobtail feeding among humpbacks would look recognizable to anyone familiar with Rogers’ framework, a slow start as a few innovators pick it up, then acceleration as it spreads through the network.