Cheap failures, early and often
The maxim is about speed, and it only works when each failure is small enough to learn from and walk away.
Fail often so you can succeed sooner.
At IDEO, the Palo Alto design firm where Tom Kelley has long been general manager, the first models of a product are often made of foam and cardboard. They are meant to be wrong. A mock-up built in an afternoon that shows a handle is awkward to hold has done its job, and it cost almost nothing.
The phrase is one of the firm’s working maxims. Kelley presents it that way in The Art of Innovation, the 2001 book he wrote with Jonathan Littman about how IDEO works, and people at the firm also quote a close variant credited to its founder, his brother David Kelley. It is sometimes passed around without a name at all.
The sentence is easy to misread as praise for failure. The goal is in the second half: sooner. Each failed prototype answers a question, and a team that answers questions ten times a week reaches a good design long before a team that answers one a month. Frequency is simply how the speed shows up.
That only works under conditions the maxim does not state. The failures have to be cheap, fast and informative. A cardboard model meets all three. So does a landing page shown to a few hundred visitors, or a paper sketch put in front of a customer. Each can be thrown away without regret.
It is poor advice where failure is expensive or cannot be undone. A medical device, an aircraft part or a bank’s ledger cannot be allowed to fail often in use; those fields fail on purpose only in labs and simulations. A founder who treats customers as test subjects can also wear out their trust, since every visible failure costs a little goodwill that is slow to earn back.
There is a subtler misuse too. A team can invoke the phrase to skip thinking, launching experiments that would have been ruled out by an hour at a whiteboard. Cheap tests are not free, and a test with no question behind it teaches nothing.
For founders the practical version is to separate experiments from commitments. Fail often, and on purpose, at the experimental stage. Before anything becomes a promise to a customer, a hire or a lender, the failing should mostly be over.