#109 Failure Has a Speed Limit

Everyone in our industry can recite the line: fail more, fail faster, failure is the tuition fee for success. We put it on slide three of the culture deck, then go back to running businesses where a single real attempt takes eighteen months to produce an answer.

The cliché isn’t wrong. It’s incomplete, and the missing part is arithmetic. Most senior people made peace with failing years ago. The question nobody asks in the boardroom is how far apart the failures are, because that number, not your tolerance for pain, decides whether you ever reach the success the cliché promises.

The Maths Nobody Runs

Say a meaningful outcome (a market that works, a product that sticks, a brand people remember) takes roughly eight serious attempts. Not a precise figure, just a stand-in for what most of us recognise: the first few goes rarely land.

Now put a clock on it. At two years per verdict, eight attempts is sixteen years. Most CEO tenures are shorter than that; most runways are far shorter. At quarterly cycles, the same eight attempts take two years. Fortnightly, four months. Same hit rate. Same skill. Completely different career.

We of all people should understand this. It’s the bookmaker’s business model. A book with a five percent margin can lose the year on ten bets: two big winners and the edge never gets the chance to show. On ten million, it won’t. Your judgement works the same way. Run it through three attempts a decade and variance drowns it. Good operators look unlucky for years for no reason other than sample size.

Bezos has said Amazon’s success is a function of how many experiments it runs per year, per month, per week, per day. Note what he didn’t say: how good the experiments were.

What Stretches the Cycle

Amy Edmondson, the Harvard Business School professor who studies how organisations learn, separates failures that teach from failures that merely cost. An intelligent failure happens in new territory, chases a clear goal, tests a hypothesis, and is as small as it can be while still telling you something. That last condition is where our industry trips. Small bets fail quickly, big bets fail slowly, and we have a cultural weakness for big bets.

Four things inflate cycle time in most operators. Bet size: a full market entry with licence, localised product and launch campaign can take eighteen months just to tell you whether it worked. Approval: every layer between idea and market adds weeks, and none of them adds information. Refusing to call it: “let’s give it another quarter” is sunk cost wearing a calendar. The verdict was available in month four; we chose not to read it.

The fourth is the one nobody puts on a Gantt chart: recovery. Failure is physiological (the stress spike, the rumination, the fortnight of quietly avoiding the topic in meetings). If it takes you a quarter to get over a miss, you’ve added a quarter to every cycle you will ever run. That’s why resilience matters. Not because it feels better. Because it’s faster.

Design for Time to Verdict

If cycle length is the variable, design for it rather than inherit it.

Size every bet so it can fail inside a quarter. If it can’t, it isn’t one experiment, it’s several stapled together, and you learn nothing until all of them finish. Set the kill criteria before launch: what number, by what date, means stop. Deciding afterwards is how “another quarter” gets approved. Keep operations and experiments in separate ledgers. Bezos’s line is that a failed 151st fulfilment centre isn’t an experiment, it’s just a failure; the same goes for a sloppy CRM migration dressed up as learning. And report verdicts reached to the board, alongside NGR and CAC. A company that reached twelve verdicts this year knows more than one that launched twelve things and is still waiting.

I’ve written before that success goes to whoever stays in the game longest. Still true. But the horizon is fixed, and what you control is how many complete cycles fit inside it. Patience with the outcome. Impatience with the loop.

So the question isn’t how many things you launched this year. It’s how many verdicts you reached. Not initiatives, not roadmaps, not “still ramping.” Verdicts.

If the honest answer is two or three, you’re not failing too much. You’re failing too slowly.