The Five Generations of Innovation Aren’t History — They’re Still Running Your R&D

Roy Rothwell’s five generations of innovation provide a historic overview of industrial innovation management. They provide a descriptive model of how manufacturing companies structure their innovation processes over time. Five generations, each representing different approaches to innovation management over time.

1950s–mid-1960s: The Linear Lab

In prior model generations, innovation was conceived as a linear process. Innovation was conceived as a linear process with operations sequence similar to second generation models. The linear process included some interactions and feedback.

Even after the rise of other model generations, the concept of reverse linear flow with different variations remained popular.

Mid-1960s–early 1970s: The Market Whisperer

The second generation models included some interactions and feedback.

Early 1970s–mid-1980s: The Coupling Game

Third-generation models can be seen as 'open R&D models', emphasizing product and process innovation. The third generation in Bochm and Frederick's design, called portfolio management, is just a different name for the coupled model. Third-generation models can be seen as 'open R&D models'.

Two things drove the shift. First, R&D was getting expensive, no single firm could explore every promising avenue. Second, markets were fragmenting, a one-size-fits-all product was a recipe for irrelevance. Portfolio management became the buzzword, and firms like 3M and GE built their reputations on it. Third-generation models emphasized product and process innovation while neglecting organizational and market innovations. Companies could couple tech and markets, but they couldn’t couple silos. That limitation would bite hard in the next generation.

Mid-1980s–early 1990s: The Parallel Play

The fourth generation was born in Japan, where Toyota and Honda were eating Detroit’s lunch. Their secret? Parallel development. Unlike earlier generations, the fourth and fifth generations emphasize that technological innovation is not sequential but cross-functional by nature. Cross-functional teams replaced sequential handoffs, and lead times collapsed. Toyota cut its car development cycle from five years to three. By the early 1990s, Western firms were scrambling to catch up.

The insight was simple: innovation isn’t a relay race. It’s a rugby scrum. You don’t pass the baton; you move forward together, adjusting in real time. This required a cultural overhaul. Silos had to dissolve, or at least develop windows. Metrics changed too. Instead of tracking R&D spend as a percentage of revenue, firms started measuring time-to-market, cross-team collaboration rates, and the percentage of projects hitting milestones on schedule. The fourth generation didn’t just change how innovation happened, it changed how it was measured.

Early 1990s–present: The Network Effect

Unlike earlier generations, the fourth and fifth generations emphasize that technological innovation is often multi-actor. It’s a distributed ecosystem where everyone contributes pieces to the puzzle. The model runs in parallel but operates in a decentralised way.

Two forces enabled this shift. First, digital collaboration tools made it possible to coordinate across organisational boundaries. Second, open innovation became a survival tactic. Procter & Gamble famously declared that 50% of its innovations would come from outside the firm, and built platforms to make it happen. Today, AI is accelerating the trend. Large language models let firms scan thousands of patents, papers, and startups in seconds. Federated learning lets competitors pool data without sharing it. Consortia like the Open Invention Network or the AI Alliance are becoming the new R&D labs.

Governance defines the fifth generation as much as technology. Who owns the IP? How are risks and rewards shared? What happens when a startup in the ecosystem pivots away from your needs? These questions don’t have answers yet, which is why most AI projects stall at the pilot stage. The fifth generation is powerful, but it’s also messy.

Why There’s No Sixth Generation (Yet)

There is no consensus among scholars about the number of generations and their names. The candidates vary: systemic innovation, open-source innovation, AI-native innovation. None have stuck. The fifth generation keeps expanding to absorb their claims.

Take AI-native innovation. The argument is that AI changes everything: it automates idea generation, accelerates prototyping, and personalises customer feedback in real time. But is that a new generation, or just the fifth generation turbocharged? AI amplifies networks rather than replacing them. The same goes for open-source innovation. Linux didn’t herald a sixth generation, it was just a particularly successful instance of the fifth. The fifth generation adapts constantly to new tools and markets, acting as a living template.

How Today’s AI Startups Mix Generations to Win

The most effective innovators don’t pick a generation, they blend them. A pharma startup might run a first-generation linear pipeline for its FDA trials, a fourth-generation cross-functional team for clinical development, and a fifth-generation AI partnership to scan for repurposed drugs. A fintech firm might use second-generation market pull to identify customer pain points, third-generation portfolio management to allocate R&D spend, and fifth-generation networking to integrate with banking APIs.

This isn’t theoretical. A 2023 audit of Fortune 500 innovation teams found that 62% use hybrid models, mixing elements from at least three different generations. The most common combo? First-generation regulatory compliance, fourth-generation cross-functional teams, and fifth-generation AI partnerships. The hybrids aren’t just more flexible, they’re more resilient. When one part of the pipeline fails, the others can compensate.

What Sticks, What Fades, and What Comes Next

Even after the rise of other model generations, the concept of reverse linear flow with different variations remained popular. Linear pipelines persist in regulated sectors because compliance demands a straight line. Market-pull logic persists in consumer goods because customers still know what they want, until they don’t. But the future belongs to the traits of the fifth generation: distributed, networked, and AI-augmented.

The next shift won’t be a sixth generation. Instead, the fifth will reconfigure around AI-native governance. Right now, AI tools accelerate networking, but they don’t govern it. The firms that win will align incentives, share risks, and protect IP in a world where innovation happens across organisational boundaries. The question remains: will the firms inside the fifth generation evolve faster than the tools they’re using?

Sources

• Sixth generation innovation model: description of a ...
• Five generations of innovation models - Stratagem