Who Gets to Tell a Failure Story?

Who gets to tell a failure story? In June 2008, J.K. Rowling stood in front of Harvard's graduating class and told one. She described the years before the books as a period in which she was "as poor as it is possible to be in modern Britain without being homeless", and she told the graduates that "failure gave me an inner security that I had never attained by passing examinations". It is the most quoted failure story of the century. harvard-rowling

Read her first sentence again. Three words in it are load-bearing, and none of them are about her character. "In modern Britain" is a description of a floor. There was something underneath the worst year of her life. She names it in the same breath as the poverty, and almost nobody quotes that half.

This is not a complaint about storytellers. The selection happens long before anyone stands up to speak. Permission to fail is a thing somebody builds, with a size, an owner and an expiry date. Where it exists, failure is survivable and the survivors talk. Where it does not, failure finishes people, and the people it finished are not in the genre at all.

Twenty-two Against Twenty-one

Paul Gompers, Anna Kovner, Josh Lerner and David Scharfstein tracked 8,753 venture-backed entrepreneurs across 9,790 ventures begun between 1986 and 2000, and counted who reached a public exit by December 2007. First-time founders: 21%. Founders who had already succeeded once: 30%. Founders who had already failed: 22%. All that scar tissue is worth a single percentage point. The sample is narrow and I will say so before anyone else does. These are venture-backed founders reaching an IPO, a population in which access to capital does an enormous amount of the work, and that objection is a real one. Hold on to it. gompers-persistence

Now the other set of numbers. John Armour and Douglas Cumming, working across fifteen countries, found that when the Netherlands introduced a three-year bankruptcy discharge in 1997, self-employment rose by roughly 4.3%, and when Germany introduced a seven-year discharge in 1999, by roughly 4.5%. Robert Fairlie, Kanika Kapur and Susan Gates found that business ownership in the United States jumps from 24.6% just under the age of 65 to 28.0% just over it, at the exact point health cover stops depending on an employer, with no comparable jump anywhere between 55 and 75. Nobody turns 65 and becomes braver. Something underneath them moved. armour-cumming fairlie-health

Somebody Wrote Down How Much Failure Was Allowed

The most honest sentence anyone has written about failure is in a Google engineering manual, and it has a number in it. Subtract the reliability target from one hundred per cent and what remains is what the book calls "the 'budget' of how much 'unreliability' is remaining" for the quarter. Product management sets that figure, not the engineers who will spend it. It is measured by "a neutral third party: our monitoring system", because a team asked to self-report its own failures will report the wrong number. And when the budget runs out, releases stop until the system is repaired. Notice who that mechanism actually binds. It is not there to make engineers brave. It is there to stop the people above them withdrawing the permission in the quarter it becomes expensive. google-sre

The same shape turns up wherever the freedom is real rather than announced. A pilot who files with NASA's Aviation Safety Reporting System gets immunity from enforcement on four conditions, one of which is that no action has been taken against them in the previous five years: permission that is rationed, dated and conditional, and therefore worth something. Etsy's engineers published their version in 2012, borrowing a distinction from safety research: a first story blames an action, a second story asks what made that action make sense to a competent person at the time. Each of these is a document. None of them is a disposition. nasa-asrs etsy-allspaw

Nobody Calls the Front of a Car Brave

A car is built to fail. Not everywhere, and not at any rate: the front of it is engineered to collapse at a chosen point, in a chosen order, at a chosen speed, so that the passenger cell does not. Whether someone walks away from a crash is settled years earlier, in a room where somebody decided where the metal would be told to give. The driver's relationship with risk has nothing to do with it. Neither does how much they have learned from previous crashes.

I Have Used the Sentence Myself

I have used the sentence myself. It did not work. I kept the fee and I kept the relationship, and the following year I was back in the same building. Someone else in the room carried the other half. I told them it was a learning experience. I want to write that I understood the problem with that sentence while I was saying it. I did not. It was true about the company. The company had somewhere to put the loss, and I never checked whether it was true about anybody else in the room.

The Strongest Case Against This, Properly Stated

Adversity does build judgment, and the argument above is close to insulting about it. A founder who has watched a company die learns things that cannot be taught in advance: which warning signs are real, which investor enthusiasm is worth nothing, how long a team keeps working after they have privately stopped believing. The research on growth after setback is not empty. And a study of venture-backed founders chasing an IPO measures the memory of the funding market at least as much as it measures anything happening inside a person; on that reading, 22 against 21 tells you who investors are willing to back twice, not who learned what. Cover is obviously a precondition for any of this. A precondition is not a cause. An account that reduces hard-won judgment to an insurance schedule has mistaken the floor for the building standing on it.

Here is where that runs out of road. Pierre Azoulay and Wesley Greenblatt, in a National Bureau of Economic Research working paper that has not been peer-reviewed, study 103,164 National Institutes of Health grant cycles between 1980 and 2015. They pull apart two things the argument above holds together: novelty and risk. Reviewers reward novelty. They penalise risk-taking, with renewal rates in the highest-risk decile lower by between 9.5 and 24.4 percentage points, and the penalty falls hardest on early-career investigators, who have the least to fall back on. A system that simply lacked nerve would punish both. This one punishes uncertainty alone, and punishes it most precisely where the floor is thinnest, which is what an account built on cover predicts and an account built on character cannot explain. azoulay-nber

Who Gets to Tell a Failure Story

The ones whose failure was absorbed by something built in advance, and who were left intact enough afterwards to narrate it. Which means that every time we coach somebody to be braver about failing, we are handing them the only part of the machine they do not control and staying quiet about the rest. They were not short of courage; they were short of cover, and I am the one who handed over a sentence that made those sound like the same thing.