For most of the twentieth century, serious R&D meant a fortress: a big internal lab, a wall around it, and a belief that the company that invented the most would win. Then the maths stopped working. Talent became mobile, venture capital learned to fund the engineer who walked out the door, and the half-life of any single secret got shorter. Open innovation is the response, a way of running R&D that treats the boundary of the firm as a membrane to manage, not a wall to defend.

The quick version

  • Open innovation means deliberately moving knowledge in (licensing, partners, users, startups) and out (spin-outs, licensing your shelved ideas), not just inventing everything yourself. The term is Henry Chesbrough's, from his 2003 book.
  • It is a complement to internal R&D, not a replacement. Procter & Gamble's "Connect + Develop" lifted the external share of its innovations from roughly 15% to over 35%, while still running a large internal lab.
  • The catch: you can only absorb outside ideas if you have enough internal expertise to recognise and use them, what scholars call absorptive capacity (Cohen & Levinthal, 1990). Cut R&D to zero and "open" innovation closes.
  • The first practical move is cheap: run an honest audit of what your team is not doing because "we don't build that here," and what valuable ideas are sitting unused on your own shelf.

The idea in depth

Henry Chesbrough, then at Berkeley, named the shift in his 2003 book Open Innovation: The New Imperative for Creating and Profiting from Technology, and laid out the argument the same year in MIT Sloan Management Review. His later, tighter definition (with Marcel Bogers, 2014) is the one worth memorising: open innovation is "a distributed innovation process based on purposively managed knowledge flows across organizational boundaries." Two words carry the weight. Distributed: good ideas are spread across many firms, universities and users, not concentrated in yours. Purposively managed: this is a discipline you run on purpose, not osmosis you hope for.

The plainest statement of why comes from Chesbrough's own framing, often summarised as a single uncomfortable sentence: not all the smart people work for you. Once you accept that, the old closed model, invent it here or it doesn't count, looks less like rigour and more like a tax you pay on your own ego.

"Not all the smart people work for us.", the premise behind open innovation

So the move is: stop measuring your R&D team only on what it invents, and start measuring it on what it ships, wherever the idea came from. A licensed component that gets you to market a year early is a win, not a confession. Rewrite the success metric before you rewrite the strategy.

Inbound, outbound, and the part everyone forgets

Open innovation has two directions, and most companies only run one. Inbound (or "outside-in") is the famous half: bringing external knowledge in through licensing, partnerships, supplier co-development, startup scouting, challenge platforms, and users. Outbound ("inside-out") is the half that gets skipped: taking ideas your business model can't use and pushing them out, licensing dormant patents, spinning out a team, open-sourcing a tool, so they create value somewhere rather than dying in a drawer.

flowchart LR
  U(["Universities & startups"]) --> F
  P(["Partners & suppliers"]) --> F
  US(["Lead users"]) --> F
  F(["Your R&D & business model"]) --> M(["Your market"])
  F --> L(["License out / spin out"])
  F --> OS(["Open-source / new market"])
Open innovation as a managed membrane: knowledge flows in from outside, and unused ideas flow back out to create value elsewhere. Leaders Loop

One source of inbound ideas deserves its own name. Eric von Hippel's research, gathered in Democratizing Innovation (MIT Press, 2005), showed that a great deal of useful innovation is done by lead users, customers at the leading edge of a need who modify or build their own solutions long before a vendor does, from surgical kit to mountain bikes to software. They are a free R&D function you are probably ignoring.

So the move is: pick one product and ask two questions this quarter. Who hacks our product to do something we never designed? (those are your lead users, go talk to them). And what have we patented or prototyped and shelved? (that is your outbound pipeline, someone may pay to license it). Both are nearly free to investigate and most firms do neither.

The honest limitation: open does not mean cheap

Here is where breathless write-ups mislead. Open innovation is not a way to fire your scientists and buy ideas off the shelf. The constraint was identified before the term existed, by Wesley Cohen and Daniel Levinthal in their 1990 paper on absorptive capacity (Administrative Science Quarterly, 35(1): 128–152). Their finding is durable and inconvenient: a firm's ability to recognise the value of external knowledge, take it in, and turn it into product is largely a by-product of its own R&D. You need internal expertise to know which outside idea is gold and which is fool's gold, and to make it work once you have it.

That is why P&G, the textbook open-innovation case, never gutted its labs. It is also why "open innovation" can quietly fail: a company guts internal capability, declares itself "open," and then can't tell a good external idea from a bad one or integrate the good ones when it finds them. Openness is a multiplier on internal capability, not a substitute for it. So the move is: before you stand up a scouting team or a challenge platform, make sure you have engineers who can evaluate and integrate what it brings back. Sourcing without absorption is just expensive tourism.

A worked example

The cleanest real case is Procter & Gamble's Connect + Develop programme, documented by P&G's Larry Huston and Nabil Sakkab in the Harvard Business Review (March 2006). Around 2000, CEO A.G. Lafley concluded that spending ever more on internal R&D for ever smaller returns was a dead end, and set a deliberate target: source a large share of innovation from outside the company. By the time the article was written, the external-sourced share of P&G's innovations had risen from roughly 15% to more than 35%, and, by the authors' account, the success rate of innovation had more than doubled while the cost of it fell. (Treat the headline figures as P&G's own reported numbers, not independent audit.)

Now make it your size. Imagine a 40-person software firm whose roadmap is jammed: a fraud-detection feature customers keep asking for, but no one in-house has built one. The closed instinct is to hire two ML engineers and spend a year. The open move, in order: (1) scout, is there a startup or open-source library that already does 80% of this? (2) absorb, can our existing engineers evaluate and integrate it (absorptive capacity)? (3) partner or license rather than build the commodity part; (4) point your own scarce R&D at the 20% that is genuinely yours. The figures here are illustrative, but the sequence is the discipline: build only what is differentiating; source the rest.

flowchart TD
  A(["Need on the roadmap"]) --> B{"Does it differentiate us?"}
  B -- "No" --> C(["Scout: license / partner / open-source"])
  B -- "Yes" --> D(["Build in-house"])
  C --> E{"Can we absorb & integrate it?"}
  E -- "Yes" --> F(["Ship faster, cheaper"])
  E -- "No" --> G(["Build internal capability first"])
A build-vs-source decision the open way: differentiation decides what you build; absorptive capacity decides what you can safely source. Leaders Loop

Frequently asked questions

Isn't open innovation just outsourcing R&D?

No. Outsourcing hands a defined problem to a contractor and gets a deliverable back. Open innovation is about knowledge flows in both directions across your boundary, pulling in ideas you didn't specify, and pushing out ideas you can't use. And it depends on keeping strong internal R&D to absorb what comes in. Pure outsourcing erodes exactly the capability open innovation relies on.

Doesn't being "open" mean giving away our edge?

Managed openness is selective, not naïve. You build in the open on the commodity layers and keep your genuinely differentiating work close. Chesbrough's point is that for most of what a company touches, secrecy buys little, the knowledge is already distributed, so the advantage shifts to who integrates and ships fastest, not who hoards.

We're small. Is this only for companies with P&G's budget?

The opposite. A large firm can afford to invent broadly; a small one can't, which makes "build only the differentiating 20%, source the rest" a survival strategy rather than a luxury. Open-source libraries, API partnerships and lead-user feedback are mostly free. What it costs you is the discipline to stop building things you could borrow.

What's the most common way it fails?

Cutting internal capability in the name of openness, then being unable to evaluate or integrate external ideas, the absorptive-capacity trap. The second most common failure is running only the inbound half and never harvesting value from your own shelved ideas.

Where does generative AI fit?

AI lowers the cost of scanning the distributed landscape (papers, patents, repos, communities) and of prototyping a borrowed idea quickly, so it strengthens the inbound side. It does not remove the absorptive-capacity constraint: you still need humans who can judge whether the idea is sound and worth integrating.

Related in the Toolkit

Where to go next