Small bets, honestly kept

There's a kind of project everyone recognises: the big one. Six months, three teams, a launch date announced at the all-hands. It feels professional. It feels like ambition. Statistically, it feels like that right up until it quietly becomes next year's big one, relaunched with a new name.

The alternative is embarrassingly modest: small bets. Work sized so that a miss costs you weeks, not seasons. A test you can run before the quarterly review, not after the reorganisation. The counter-intuitive part is that small bets aren't less ambitious, they're ambition with a seatbelt. Ten honest experiments will out-learn one heroic project almost every time, and the compounding is absurd.

The hard part isn't starting small. It's staying small. Every stakeholder wants their feature to be The Feature. Every engineer can see the elegant generalisation. Every PM feels the pull to make the thing properly, which in practice means bigger. The discipline is in writing down, upfront, the smallest version of the bet that would still teach you something, and refusing to fund anything larger until that version has told you the truth.

I've started calling this “keeping bets honest”. A bet is honest when the team could plausibly be glad it failed, because they'd have learned something cheap that redirects the next three bets. If failure would be a catastrophe, the bet isn't a bet. It's a hope with a budget line.

Big plans make for better keynotes. Small bets make for better products, and (more to the point) better Tuesdays.