Fast-twitch muscle
Disruptive experimentation
Fast-cycle hypothesis testing.
Less about reckless bets, more about systematic digging. Prototype, pilot, and challenge the assumptions underneath the legacy model. The question is always: what do we need to learn, and what is the cheapest way to learn it?
How you train it.
Reward intelligent failure and fast learning loops, not just launches. Build, measure, learn, on a rhythm, and keep the experiments small enough to stay experiments.
On your own: Try the small version of the thing you have been putting off, this week, and treat whatever happens as information.
Who does it well.
A year perfecting a product nobody wanted. A one-day pop-up test would have answered the question first.
Over a thousand simultaneous A/B tests. Experimentation as collective discipline, not individual heroics.
Drills.
The improviser's rule: accept what your partner offers and build on it, before you judge it. Run a meeting that way once and watch how many ideas survive their first minute.
Improvisational theater; Kelly Leonard and Tom Yorton, Yes, And (Second City)
Sort failures into three kinds: basic (avoidable), complex (systems) and intelligent (a small, well-designed try in new territory). Punish none, and reward the third kind out loud. People don't experiment where failure is unspeakable.
Amy Edmondson, Right Kind of Wrong
State the riskiest assumption, build the smallest thing that tests it, measure, decide. The loop matters more than any single experiment.
Eric Ries, The Lean Startup
List every assumption behind the idea, rank by importance and by how little evidence you have, and test the top right corner first.
David Bland and Alexander Osterwalder, Testing Business Ideas
Fake the product before you build it: a landing page, a concierge version, a button that does nothing yet. Answers whether anyone wants it before whether you can make it.
Alberto Savoia, The Right It
A/B testing done properly, with sample sizes, guardrail metrics and a culture that accepts most tests will lose.
Ron Kohavi, Diane Tang and Ya Xu, Trustworthy Online Controlled Experiments
Write the press release and FAQ for the finished thing first. If the release is dull, don't build it; if it's exciting, you now know what to test.
Amazon's PR/FAQ; Colin Bryar and Bill Carr, Working Backwards