Sample frameDay 184·Innovation · Search

Good morning. Nobody in your business decided to stop looking for the next thing. It simply stopped being worth anyone's while.

Innovation.

Exploration and Exploitation

5 min read·Apply by lunch

The question

Add up where your best people went last quarter: how many hours went into making something you already sell work slightly better, and how many went into finding out whether something you don't sell yet could work at all — and if the second number is close to zero, notice that nobody chose it. Every individual decision that produced it was correct, which is exactly what makes it so hard to reverse.

The idea

James March spent the long back half of his career at Stanford, holding chairs in management, political science, sociology and education at the same time, and in 1991 he published a paper in Organization Science called Exploration and Exploitation in Organizational Learning. Its argument is arithmetic rather than moral. Exploitation — refining, selecting, executing, making the existing thing better — has returns March described as positive, proximate and predictable. Exploration — search, variation, experiment, discovery — has returns he described as uncertain, distant and often negative. If that sounds like Christensen's innovator's dilemma, it isn't the same claim. Christensen explains why an incumbent rationally declines one particular low-margin market, and he needs a disruptor for the story to bite. March's result is more general and a good deal more uncomfortable: the same arithmetic strips search out of any organisation's portfolio with no disruptor anywhere in sight, and because organisations learn from their own experience, they get better at whichever one they do more of — which raises exploitation's returns again and makes the next comparison more lopsided still. Nobody in the building is being short-sighted. They're each correctly preferring a certain return to an uncertain one, and the sum of correct preferences is a business that has quietly stopped looking.

The framework

01
Every fair comparison you run will favour the thing that already works.
This is the mechanism, and it isn't corrected by wishing it corrected. Put a proposal to improve the existing product beside a proposal to test an unproven one, ask both for a return and a date, and the first wins — not because your process is broken but because it's working exactly as designed. Three things separate them: certainty, time and proximity. Any rule that ranks by expected return will strip search out of the portfolio within a few cycles, and it'll do it while every individual decision still reads as sound in the minutes.
02
The slow learners.
The oddest result in the 1991 paper came from a simulation of people learning from a shared organisational code while the code learned back from them. When people absorbed the house view quickly, the organisation converged fast on what it already believed, and its knowledge in the long run was lower — everyone agreed early, so nobody was left holding the belief that turned out to be right. Slower learners, and a steady trickle of new arrivals who hadn't yet adopted the code, kept alternatives alive long enough for some to pay off. The variety you can't generate internally is already on the payroll, and it has a shelf life.
03
Stop asking the search what it will return. Ask it what it will settle.
You can't make exploration win on returns, so take it out of that comparison and give it a different question to answer. A search project's output isn't revenue; it's a reduction in the uncertainty around a belief you're currently betting on without evidence. Judge it that way — what did we believe on Monday, what do we know on Friday, what would we now refuse to fund. Written like that, a test that kills an idea is a full return rather than a write-off, which is the only accounting under which anyone volunteers to run one.
04
Which of your working answers has never been re-opened?
Exploitation's returns stay certain right up until the world moves, and the better you've got at the current answer, the longer you'll keep collecting on it while the ground shifts underneath. So go looking for the practices you've refined for years without ever re-testing the assumption beneath them: the channel that always worked, the customer type you always sold to, the price you set once and indexed ever since. Competence is what makes them invisible. Nobody re-opens a question that's been answered profitably for years, and that's exactly the class of question that goes stale without announcing it.

Members only · 4 principles + template + AI mentor

364 more frameworks are waiting.

Exploration and Exploitation is Day 1 of 365. One framework every morning for a year — across strategy, sales, negotiation, leadership, and more.

$1 /day

Billed $365/yr · cancel any time

365 frameworks

12 topics · templates · AI mentor

Reserve your place — $365/yr →

One payment a year · the whole library · no auto-renewal.

More Innovation frameworks

All 24 innovation frameworks →