The Explore-Exploit Toggle
Exploration reliably precedes a hot streak — then you drill, then you zoom back out
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 92%
The fundamental tension Epstein names for successful individuals and organisations: explore means looking for new knowledge or capabilities that might add value; exploit means drilling down on what you already do well. Once people or firms find something they're good at, they tend to abandon explore mode entirely and settle at a middling level. Dashun Wang's analysis of roughly 26,000 film directors, artists and scientists found careers come in hot streaks, most people get only one, and what reliably precedes a hot streak is a period of exploration — trying different styles, going broad, staying nimble, moving between teams — followed by drilling in. Repeat performers zoom back out and explore again.
Origin
Epstein raises it in response to Bartlett describing hiring a head of failure at Diary of a CEO, and grounds it in Dashun Wang's large-sample career studies at Northwestern.
Core principles
- 01Success creates a disincentive to change anything you're doing.
- 02Exploration precedes hot streaks; skipping it caps you at a middling exploit level.
- 03You want to be in the zone of optimal push, failing 15-20% of the time.
- 04A rut of competence is comfortable enough to feel like a hammock.
How to run it
- 1
Notice you're in exploit mode
Recognise when competence and positive feedback have made changing anything feel unnecessary or risky.
Pro tip Russ Roberts' phrase for it is a hammock of competence, not just a rut.
Watch out Risk for a successful leader is risk for everyone who depends on them, which makes the trap self-reinforcing.
- 2
Set up low-stakes practice
Design small, cheap experiments that test what might become a worthwhile larger experiment.
Pro tip Epstein: how can you set up some low stakes practice for what might be a worthwhile larger experiment.
- 3
Calibrate to the zone of optimal push
Push hard enough that you're failing 15-20% of the time, which is the rate that maximises improvement.
Watch out A metric going backwards during an experiment is not a failure of the experiment.
- 4
Toggle, don't settle
After exploration surfaces something, drill into it hard — then deliberately zoom back out to explore again for the next streak.
Watch out Staying permanently in one mode is what sacrifices the hot streak.
In the wild
When the podcast hit number one in Europe, Bartlett hired someone whose sole responsibility was increasing the rate of failure and experimentation, then placed a failure assistant inside each of the production, social, commercial and booking teams.
→ A TikTok discovery by the social team transferred straight to the guest research team — the one plus one equals three effect of reporting failures across departments.
Kodak invented the digital camera and scuttled it, reasoning there was no case for disrupting its own business.
→ The archetypal cost of ditching explore mode once exploit is profitable.
Common mistakes
Lionising failure without underwriting it
Saying failure is good is lip service; Adam Grant's hippo effect means signals from the highest-paid person in the room decide what people actually do, so titles and visible risk matter.
Failing only trivially
Epstein notes a leader has to fail at something of consequence to set the agenda — tripping on a curb during a jog doesn't count.
Is it for you?
Best for
Successful individuals and teams whose competence has quietly stopped them experimenting.
Not ideal for
Early-stage operators still searching for anything that works, who are already all explore.
From the transcript
“the fundamental problem you're getting at here is the one called the explore exploit tradeoff”
“the clear message of his work is that exploration precedes a hot streak”
“if you're not at least like 15 20% of the time failing then you're not in your zone of optimal push”
From the episode
David Epstein: 10,000 Hours Is A Lie! The Morning Habit That’s Secretly Ruining Your Day!
David Epstein