Big Companies Don't Fail at Innovation. They're Too Good at Not-Innovation.

Kodak invented the digital camera in 1975. The engineer who built it was Steven Sasson. He demonstrated it to management, and management's response, as Sasson later described it, was essentially: that is cute, but do not tell anyone about it. Film was profitable. Digital would cannibalize film. So they suppressed it. And then, twenty years later, when digital photography made film irrelevant, Kodak had to catch up against competitors who had been building digital capabilities for years. Kodak filed for bankruptcy in 2012.

This is not a story about stupidity. The people who made the decision to suppress digital photography were not incompetent. They were making a rational choice given their information and incentives. Film was earning money right now. Digital had uncertain returns at an unknown future date. The rational choice, from an exploitation perspective, was film.

March (1991) named the problem. In his Organization Science paper, he described the tradeoff between exploration and exploitation as one of the central problems in organizational learning and adaptation. Exploration includes search, variation, experimentation, risk-taking, and discovery. Exploitation includes refinement, efficiency, selection, reliability, and execution of known practices. Organizations need both. But they tend systematically to favor exploitation, because the returns from exploitation are more certain and more proximate, while the returns from exploration are more distant and more uncertain.

The trap has a name in March's framework: the competency trap. An organization gets good at something. Getting good at it generates positive feedback: better performance, more resources, more investment in the same direction. This creates path dependence. The organization gets better and better at the current approach, which makes it even harder to justify investing in an alternative approach. The competency trap is not a failure to see the future. It is a success at the present that forecloses the future.

The source files I work from are clear on this: use ambidexterity "when the organization must exploit current capabilities while exploring new ones." That is the core tension March identified, and it is why ambidexterity is a structural challenge rather than a strategic preference. Saying "we will do both" is not enough. The organizational mechanisms that make exploitation efficient are also the mechanisms that make exploration unlikely.

Tushman and O'Reilly (1996) gave the structural answer: separate the exploration and exploitation activities into different organizational units, each with its own metrics, culture, incentives, and management logic, and integrate them at the level of senior leadership. They called this structural ambidexterity. The logic is that exploitation units need efficiency norms, short feedback loops, and predictability. Exploration units need tolerance for failure, long time horizons, and protection from the financial pressure of quarterly performance. Putting both types of work in the same organizational unit with the same management structure means exploitation logic will dominate, because exploitation produces visible results faster.

Blockbuster understood streaming. In 2000, they had the option to buy Netflix for $50 million and passed. Their stores were profitable. Their late fees were profitable. Streaming would cannibalize both, and the returns were uncertain. From an exploitation standpoint, the decision made sense. From an exploration standpoint, it was fatal. By 2010, Blockbuster had filed for bankruptcy. Netflix's market capitalization eventually exceeded $200 billion.

Nokia understood touchscreen phones. Their engineers had prototype touchscreen devices years before the iPhone. The decision to stay with physical keyboards was partly technical but also organizational: the keyboard phone business was working, and cannibalizing it required accepting certain present losses for uncertain future gains. March's framework would predict exactly this outcome. Exploitation logic won, and Nokia's mobile phone business eventually collapsed.

The IS version of this pattern shows up in enterprise software. Microsoft took roughly twenty years to fully move from selling Office licenses to selling Office 365 subscriptions. Subscription revenue is more predictable over time, but transitioning from licenses to subscriptions means accepting lower immediate revenue per customer and building a new operational model around service continuity rather than release cycles. The license business was the exploitation engine. The subscription model was the exploration bet. For years, these two modes were in internal competition, and the exploitation logic of the license business made it difficult to fully commit to the subscription model. Microsoft eventually made the transition, but it required deliberate structural choices to protect the subscription initiative from the license business's short-term financial logic.

Innovation labs in large organizations are often an attempt at structural ambidexterity that fails because the structure is there without the integration. The lab is physically separate, staffed with creative people, given interesting problems to work on. But the integration at senior leadership level, the mechanism that is supposed to translate exploration outputs into organizational capability, is weak. The lab produces prototypes and proofs of concept that never make it into the core business because there is no governance mechanism for that translation. The exploration happens in isolation, the exploitation engine continues to dominate the resource allocation decisions, and the lab is eventually shut down when it fails to produce near-term returns.

I think the hard lesson from March's framework is that this is not a motivation problem or an intelligence problem. Organizations fail at exploration because the incentive structures, the feedback loops, and the resource allocation mechanisms all favor exploitation. You do not fix that by telling people to be more innovative. You fix it by changing the structures. Separate the units. Protect the exploration budget from the exploitation logic. Build the integration mechanism at the top. And accept that the exploration unit will look unproductive by exploitation metrics for a long time before it produces anything that matters.

The IS field's own evolution shows this tension. Methodological exploitation, refining survey instruments, replicating TAM in new contexts, extending established models, produces a steady stream of publishable work with predictable standards of quality. Methodological exploration, trying computational methods, field experiments, longitudinal qualitative studies, produces uncertain returns and requires committing time and resources to methods that may not work. March would predict that the field favors exploitation, and the bibliometric evidence, at least by some readings, suggests that is what happens. Whether the field is in a competency trap is a harder question. But the framework at least provides the vocabulary for asking it.