The Explore/Exploit Balance
Keep investing in random search even when exploiting works fine
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 4
- Confidence
- 83%
The Explore/Exploit Balance is the trade-off, drawn from AI, animal foraging and Brian Christian's Algorithms to Live By, between using what you already know and searching for something better. You are an idiot if you never exploit existing knowledge, but equally foolish if you only exploit on the assumption your knowledge is complete and permanent. Bees model it: most fly to known pollen, but random bees search elsewhere, future-proofing the hive against sudden depletion and creating the chance of a bonanza. Cut the random bees and short-term nectar figures improve, then it is downhill, because you can't cost-cut your way to growth. Most businesses over-optimise for exploit once past the entrepreneurial phase, so the discipline is to protect a deliberate exploration budget even when it dents efficiency.
Origin
Rory Sutherland applies the explore/exploit trade-off from AI and foraging biology to marketing and business, citing Brian Christian's Algorithms to Live By. Extracted from Deep Dive with Ali Abdaal.
Core principles
- 01Ignoring what you already know is stupid; exploiting only what you know is fatal
- 02Exploitation looks more efficient in the short term and is downhill after that
- 03Exploration buys future-proofing against sudden depletion
- 04You cannot cost-cut your way to growth
How to run it
- 1
Measure your split
Estimate how much of your resource currently goes to exploiting known winners versus exploring the unknown.
Pro tip Mature firms are almost always over-indexed on exploit.
- 2
Ring-fence exploration
Set aside a protected budget for random, speculative search that is not expected to pay off immediately.
Pro tip Frame it as future-proofing, not waste.
Watch out It will always look inefficient on this quarter's numbers.
- 3
Protect the random bees
Resist the pressure to cut speculative activity just because removing it flatters short-term efficiency.
Watch out Killing exploration makes today's figures look better and guarantees decline afterwards.
- 4
Harvest the surprises
Treat unexpected upside discoveries as the payoff, and feed learning back into where you next explore.
Pro tip Without explore you only ever get nasty surprises, never pleasant ones.
In the wild
Sutherland's book had an obvious target audience of marketing and psychology readers. But a single unplanned appearance on the Chris Evans breakfast show reached a wildly different audience: Evans read it, raved about it, and the book briefly outsold JK Rowling on Amazon. The bonanza came from an explore channel he could never have defined in advance.
→ An unpredictable exploration path produced an outsized, undefinable win.
A hive sends most bees to known pollen but keeps random bees searching at random. If you removed them, short-term nectar figures would look more impressive, but the hive would lose its defence against depletion and its shot at a big new find.
→ Cutting exploration boosts near-term metrics and dooms long-term growth.
Common mistakes
Optimising away exploration
Removing speculative activity to improve short-term efficiency removes the only source of future breakthroughs and resilience.
Assuming knowledge is complete
Exploiting on the belief that the world will not change leaves you defenceless when it does.
Is it for you?
Best for
Businesses past the entrepreneurial phase deciding how much to spend on efficiency versus open-ended search.
Not ideal for
Genuinely stable, complete domains where the environment will not change and nothing new can be discovered.
From the transcript
“there's a trade-off between exploration and exploitation... you're a total idiot if you don't exploit what you already know... but you're equally foolish... if you…”
“you can't cost cut your way to growth”
From the episode
How To Influence People: Marketing Secrets Behind The World’s Biggest Brands - Rory Sutherland
Rory Sutherland