Everybody quotes the number. When I interviewed Sir James Dyson, he gave me it plainly: "I created 5,127 prototypes of the first Dyson bagless vacuum cleaner and only the last one was right! Not being afraid to fail is something I think all successful entrepreneurs have in common."
The number gets repeated as a story about persistence. It is really a story about record-keeping. The 5,126 that failed were only useful because somebody wrote down what each one did and why it was wrong.
Most organisations that tell me they learn from failure have no equivalent record. They have a feeling that they learned something: held collectively, written nowhere, gone within a quarter.
Failure is the one topic where the rhetoric has run miles ahead of the practice. We have all accepted that failure is instructive. Almost nobody has built the dull apparatus that turns an instructive event into an actual instruction.
01Why doesn't "we learn from failure" produce a lesson?
Because the thing people examine is the outcome, and the outcome is not where the lesson lives.
The distinction that matters is between a decision that was sound on the information available and one that ignored information already sitting in the building. Only the second is a mistake in any useful sense. Collapse the two and you teach a company that the safest thing anyone can do is decide nothing at all.
That cuts both ways. A team that reviews only its losses rebuilds processes that were fine and simply unlucky. A team that reviews only its results leaves a reckless decision unexamined because it happened to pay.
So the first question in an honest review is not what went wrong but what was known at the time, and by whom. Hindsight makes everybody look negligent, which is why it is the enemy of the exercise.
02What does a real post-mortem look like?
It looks boring, it happens on a fixed date, and it produces a document.
When I spoke to Daniel Lamarre, then chief executive of Cirque du Soleil, he described the habit rather than the sentiment: "My learnings came not only from our successes, but a lot of the key learnings came from failures... when you do, rather than to run away from it, you have to sit down and do a good post-mortem, and learn from your failures." The verb that matters there is sit down. It is scheduled work, not a mood.
The sequence I have seen work is unglamorous. Build the timeline first, in facts, before anybody is allowed to interpret. Ask what was known at each decision point. Classify: bad process, or bad luck. Then, and only then, one named owner, one specific change, one date. A review that ends without an artefact has not happened.
Sheryl Sandberg described the most memorable version of this to me, from a manager she admired: "She had this thing called, 'Whoops the Monkey!', where every single week she would have her team come together and say what the biggest mistake was that someone had made." A weekly ritual with a silly name does what a solemn annual inquest cannot. It makes reporting routine, which is the only way it becomes truthful.
There is a discipline in not over-learning, too. Michael Bisping, the UFC middleweight champion, told me what he does after a loss: "When that happens, you say, 'OK, what did I do wrong?' You have to be honest with yourself... Maybe you just got beat by someone who was better than you, it happens, there's no shame in that." Sometimes the correct output of a post-mortem is that nothing changes. Teams that cannot reach that conclusion thrash their own systems after every bad quarter.
A failure you cannot describe in writing is not a lesson. It is an anecdote with a moral attached afterwards.
03Does the language you use change what people admit to?
Yes, and more than almost any policy you could write.
The clearest example came from Owase Jeelani, the paediatric neurosurgeon at Great Ormond Street who separates conjoined twins. He explained why his field changed its vocabulary: "Words like mistake, error and complication are not helpful. They carry visceral, emotive, weight which hampers learning... we've moved away from that terminology towards the language of adverse events."
This is not softening; it is the opposite. Neutral language raises the reported volume of things going wrong, and volume is what makes a pattern legible. A word that carries shame produces silence, and silence in a surgical unit is far more dangerous than an uncomfortable meeting.
The same logic runs through commercial organisations, usually applied badly. Tony Hsieh, the founder of Zappos, told me his team had made a deliberate substitution: "I don't think failure is the right word to use... rather- experimentation. We want to move towards incorporating experimentation into our everyday language."
I would take that seriously and hold it at arm's length. Reframing a project as an experiment is legitimate when it genuinely was one, with a hypothesis and a stopping rule agreed in advance. It is evasion when applied afterwards to something that was meant to work and did not. The test: was it called an experiment before the result came in?
04Why should a run of unbroken success worry you?
Because a hundred per cent success rate is a statement about the size of your attempts.
Ed Catmull, who co-founded Pixar and ran animation at Disney, was blunt about what an absence of failure signifies: "If you don't have some failures continually, it is a signal that you have retreated into the conservative past." An organisation with nothing to review has usually not become excellent. It has narrowed its ambitions to what it can guarantee.
General Stanley McChrystal gave me the mechanism from a setting where the stakes were rather higher. "In wartime, people are much more afraid of failure than physical harm... A little bit of fear is good, it generates a creative tension and gets people to respect the task at hand, but as soon as you're more scared of failure than you are excited about succeeding, that's when fear becomes a problem."
That threshold is the one leaders control. You cannot legislate courage, but you can watch what happens to the last person who tried something ambitious and missed, because everybody else already has, and this is where Not published yetRisk and Recklessness Are Not the Same Word in BusinessScheduled for 26 October 2026. We publish two or three articles a week, so check back then or see what is published. actually gets drawn.
05The three ways a post-mortem quietly fails
Three failure modes, and only one of them is about honesty.
The review happens too early. A post-mortem held in the week of the failure captures emotion, not causation. People are defending positions. Two to four weeks later, with the timeline assembled in advance, produces a better document.
Nobody owns the change. The commonest outcome of a corporate post-mortem is shared understanding and no altered process. Six months later the same failure arrives wearing different clothes. A review that ends without a name and a date has produced comfort, not learning.
The person who decides promotions is in the room. Psychological safety is not a value, it is a seating plan. Honesty about a costly mistake is a career calculation, and people make it accurately, which is the same reason Not published yetThe Cost of a Quiet Room: Why Consensus Leads to Bad DecisionsScheduled for 11 January 2027. We publish two or three articles a week, so check back then or see what is published.. Run the factual reconstruction without the person who writes the appraisals.
06When failure really is just failure
The honest objection is that a great deal of failure is not tuition. It is just loss, and calling it learning avoids the conclusion that something was done badly by people who should have known better.
That objection holds in specific places. In safety-critical and regulated work the primary duty is prevention, and a tolerance for failure that reads as tolerance for harm is indefensible. In a business with a few months of cash, the capacity to absorb an instructive failure does not exist, and the honest answer is to stop experimenting until it does. In any operation whose value is reliability, the same mistake twice is not a learning culture but an execution problem with better vocabulary.
The distinction worth holding is between failures of attempt and failures of care. The first are the price of doing anything new and deserve the full apparatus above. The second deserve a different conversation, and confusing the two is how organisations excuse sloppiness in the language of innovation. There is also a version of this culture where failing is celebrated rather than examined, and the celebration quietly replaces the analysis.
07The whole difference is administrative
Pick the most expensive thing that did not work last year. Reconstruct the timeline in facts before anybody interprets it. Establish what was known at each decision point. Then write one page: bad process or bad luck, one change, one owner, one date. It will take an afternoon.
Reggie Fils-Aimé, who ran Nintendo of America, described what that discipline is worth at scale. The Wii U sold badly by any measure the company cared about. "The company was able to see the elements that didn't work, the elements that did work, and they created the Nintendo Switch. That is now the best-selling console Nintendo has ever had."
The Switch did not emerge from a willingness to fail. Plenty of companies have that. It emerged from somebody separating what worked from what did not and writing it down where the next team could find it. That is the whole difference between an organisation that fails and one that learns, and it is administrative far more often than it is cultural.