Exploration and Exploitation
Exploit proven winners while preserving a budget for discovery
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 99%
Exploration and exploitation is an allocation strategy for decisions under uncertainty. When little is known, sample several options and gather real-world results. As evidence accumulates, direct an increasing share of time, money, and effort toward the options that perform consistently. Crucially, do not commit 100 percent to the apparent winner. Random variation may have made a merely local maximum look globally best, while competitors or changing conditions uncover stronger alternatives. Preserve a continuing exploration budget for experiments and fresh information. Kaufman argues that this serves two purposes at once: it is a learning strategy for a new business and a long-term competitive strategy for an established one. It also discourages betting the entire company on an untested assumption.
Origin
Kaufman added the concept to the tenth-anniversary edition of The Personal MBA from research in decision theory and computer science.
Core principles
- 01Begin with broad exploration when information is scarce
- 02Shift resources toward options that repeatedly perform
- 03Never allocate everything to the current apparent winner
- 04Continued experiments protect against a local maximum
How to run it
- 1
Explore broadly
When you lack information, test several options to learn how the world responds.
Pro tip Keep early tests small enough that failure is affordable.
Watch out Do not bet the business on what is still an experiment.
- 2
Compare evidence
Gather outcome data and identify which options produce better and more consistent results.
Pro tip Use observed behavior rather than confidence or preference.
- 3
Exploit the leaders
Shift a growing share of resources toward the strongest demonstrated options.
Pro tip Increase allocation progressively as evidence strengthens.
Watch out A short run of good results may be statistical noise.
- 4
Protect exploration
Continue dedicating a fixed share of resources to alternatives and new experiments.
Pro tip Treat exploration as a permanent operating activity rather than a one-off phase.
Watch out Choosing the current best option 100 percent of the time can trap the business at a local maximum.
- 5
Reallocate repeatedly
Update the balance between exploration and exploitation as new evidence arrives.
In the wild
Kaufman asks the listener to imagine several free slot machines when one pays more than the others. With no initial information, the player samples machines at random, then increasingly favors the machine producing the best results while continuing to test alternatives.
→ The player benefits from the current leader without losing the chance to discover a better machine.
Common mistakes
Going all-in too early
Treating an untested assumption as certain can put the whole business at risk.
Ending experimentation
Allocating everything to the apparent winner prevents learning and can lock the business into a local maximum.
Is it for you?
Best for
It is best for businesses choosing among products, offers, channels, or operating methods under uncertainty.
Not ideal for
It is not appropriate where even a small experiment creates unacceptable irreversible risk.
From the transcript
“when you start with no information, the first strategy is pretty clear. You just play machines at random, and you collect some data from the…”
“the optimal strategy is never to choose what you think is the best option and do that 100% of the time.”
“always dedicating a certain amount of your time to exploration, to trying new things, gathering new data, seeing what works is absolutely critical.”
From the episode
The Key Principles Of Running Any Business - Josh Kaufman - #215
Josh Kaufman