It is an extension of the trend that Don Swanson foretold, and it massively increased opportunities for Yokoi-like connectors and polymathic innovators. “When information became more widely disseminated,” Ouderkirk told me, “it became a lot easier to be broader than a specialist, to start combining things in new ways.”
* * *
• • •
Specialization is obvious: keep going straight. Breadth is trickier to grow. A subsidiary of PricewaterhouseCoopers that studied technological innovation over a decade found that there was no statistically significant relationship between R&D spending and performance.* (Save for the bottom 10 percent of spenders, which did perform worse than their peer companies.) Seeding the soil for generalists and polymaths who integrate knowledge takes more than money. It takes opportunity.
Jayshree Seth rose to corporate scientist precisely because she was allowed to pinball around different technological domains. Staying in one technical lane isn’t her thing. Seth was unenthusiastic enough about the research she did for her master’s degree that she ignored warnings and switched labs at Clarkson University for her PhD in chemical engineering. “People said, ‘This is going to take too long because you have no fundamental knowledge in this area and you’re going to be behind people who have already done their master’s there,’” she told me. To clarify: the advice she received was to stick in an area she knew she didn’t like because she had already started, even though she wasn’t even that far in. It is the sunk cost fallacy embodied.
When she entered the professional world with 3M, she dared to switch focus again, this time away from her PhD research, and for a personal reason: her husband was coming to 3M from the same Clarkson lab, and she didn’t want to occupy the spot he might apply for. So she branched out. It worked: Seth has more than fifty patents. She helped create new pressure-sensitive adhesives for stretchable and reusable tapes, and diapers that stay on wiggly babies. She never studied materials science at all, and claimed she is “not that great a scientist.” “What I mean,” she said, “is I’m not qualified fundamentally to do what I do.” She described her approach to innovation almost like investigative journalism, except her version of shoe-leather reporting is going door-to-door among her peers. She is a “T-shaped person,” she said, one who has breadth, compared to an “I-shaped person,” who only goes deep, an analog to Dyson’s birds and frogs. “T-people like myself can happily go to the I-people with questions to create the trunk for the T,” she told me. “My inclination is to attack a problem by building a narrative. I figure out the fundamental questions to ask, and if you ask those questions of the people who actually do know their stuff, you are still exactly where you would be if you had all this other knowledge inherently. It’s mosaic building. I just keep putting those tiles together. Imagine me in a network where I didn’t have the ability to access all these people. That really wouldn’t work well.”
In his first eight years at 3M, Ouderkirk worked with more than a hundred different teams. Nobody handed him important projects, like multilayer optical film, with potential impact spanning an enormous array of technologies; his breadth helped him identify them. “If you’re working on well-defined and well-understood problems, specialists work very, very well,” he told me. “As ambiguity and uncertainty increases, which is the norm with systems problems, breadth becomes increasingly important.”
Research by Spanish business professors Eduardo Melero and Neus Palomeras backed up Ouderkirk’s idea. They analyzed fifteen years of tech patents from 32,000 teams at 880 different organizations, tracking each individual inventor as he or she moved among teams, and then tracking the impact of each invention. Melero and Palomeras measured uncertainty in each technological domain: a high-uncertainty area had a lot of patents that proved totally useless, and some blockbusters; low-uncertainty domains were characterized by linear progression with more obvious next steps and more patents that were moderately useful. In low-uncertainty domains, teams of specialists were more likely to author useful patents. In high-uncertainty domains—where the fruitful questions themselves were less obvious—teams that included individuals who had worked on a wide variety of technologies were more likely to make a splash. The higher the domain uncertainty, the more important it was to have a high-breadth team member. As with the molecular biology groups Kevin Dunbar studied that used analogical thinking to solve problems, when the going got uncertain, breadth made the difference.
* * *
• • •
Like Melero and Palomeras, Dartmouth business professor Alva Taylor and Norwegian School of Management professor Henrich Greve wanted to examine the creative impact of individual breadth, just in a slightly less technical domain: comic books.
The comic book industry afforded a well-defined era of creative explosion. From the mid-1950s to 1970, comic creators agreed to self-censor after psychiatrist Fredric Wertham convinced Congress that comics were causing children to become deviants. (Wertham manipulated or fabricated aspects of his research.) In 1971, Marvel Comics broke ranks. The U.S. Department of Health, Education, and Welfare asked Marvel editor in chief Stan Lee to create a story that educated readers about drug abuse. Lee wrote a Spider-Man narrative in which Peter Parker’s best friend overdosed on pills. The Comics Code Authority, the industry’s self-censorship body, did not approve. Marvel published anyway. It was received so well that censorship standards were immediately relaxed, and the creative floodgates swung open. Comic creators developed superheroes with complex emotional problems; Maus became the first graphic novel to win a Pulitzer Prize; the avante-garde Love and Rockets created an ethnically diverse cast that aged with readers in real time.
Taylor and Greve tracked individual creators’ careers and analyzed the commercial value of thousands of comic books from 234 publishers since that time. Each comic required the integration, by one or multiple creators, of narrative, dialogue, art, and layout design. The research duo made predictions about what would improve the average value of comics produced by an individual or team creator, and what would increase the value variance—that is, the chance that a creator would make a comic book that either failed spectacularly compared to their typical work, or that succeeded tremendously beyond their norm.
Taylor and Greve expected a typical industrial production learning curve: creators learn by repetition, so creators making more comics in a given span of time would make better ones on average. They were wrong. Also, as had been shown in industrial production, they guessed that the more resources a publisher had, the better its creators’ average product would be. Wrong. And they made the very intuitive prediction that as creators’ years of experience in the industry increased, they would make better comics on average. Wrong again.
A high-repetition workload negatively impacted performance. Years of experience had no impact at all. If not experience, repetition, or resources, what helped creators make better comics on average and innovate?