Making the Most of Your Mistakes

2026-08-03 - source - Read full transcript

Shankar Vedantam (host)Amy EdmondsonSusan Prescott

Key insights

Demanding zero errors doesn't reduce errors, it just stops people from reporting them.

Edmondson's hospital research found teams with better interpersonal climates showed HIGHER recorded error rates, the opposite of her hypothesis. The reason: her survey item on whether mistakes were 'held against you' predicted reporting, not occurrence. Punitive cultures don't eliminate mistakes, they just make them invisible until they cause real harm.

psychological-safety

Not all failures are the same, and treating them identically is itself a mistake.

Edmondson argues for three categories: basic failures (single preventable slips, should be stamped out), complex failures (multiple small factors lining up, should be caught via systems and speaking up), and intelligent failures (hypothesis-driven experiments in new territory, should be encouraged). A blanket 'no failure tolerated' or blanket 'fail fast' culture misapplies the wrong response to the wrong category.

failure-taxonomy

The context determines whether a fail-fast or zero-defect culture is appropriate.

'Fail fast, fail often' works for an R&D lab exploring new territory, but is the wrong approach for an operating theater, an airline cockpit, or an assembly line, where the goal is disciplined execution of known procedures, not experimentation.

failure-taxonomy

Complex failures result from multiple small factors lining up, not one root cause.

Using James Reason's Swiss cheese model, Edmondson traced a child's morphine overdose to seven contributing factors: an ICU overflow that sent a post-surgical patient to a less-specialized floor, a brand-new nurse, a poorly placed infusion pump, an unverified independent calculation, and a badly printed medication label. No single factor caused the harm; their alignment did.

failure-taxonomy

Catching small problems early is far cheaper than fixing large ones downstream.

Toyota's Andon cord lets any line worker flag a possible issue; a leader investigates within moments, and only about 1 in 12 pulls actually stops the line. Each stoppage costs one car sale, but the alternative, letting a defect propagate, is far more expensive. Toyota treats the cost of stopping as an investment, not a loss.

risk-management

Basic failures often happen in moments of false confidence, when a task feels routine.

Edmondson's own sailing accident (getting knocked overboard by a boom while relaxed and inattentive near the dock) and the Air Florida Flight 90 crash (a checklist item skipped 'habitually') both illustrate that basic failures cluster around tasks people believe they can do 'in their sleep.' The complacency itself is the risk factor.

risk-management

Intelligent failures require three things: they occur in new territory, are driven by a hypothesis, and are kept as small as possible.

Chemist Jennifer Heemstra tells her lab 'we're going to fail all day' because roughly 90% of well-informed hypotheses in frontier research don't pan out; each failure narrows the search space. Edmondson's own mother applied the same logic in dating: a bad blind-date weekend led her to agree only to a drink for the next introduction, bounding the potential downside before it turned out to be her future husband.

decision-making

Doing your homework before attempting something new is what separates an intelligent failure from a reckless one.

Whether it's a scientist reviewing the literature before running an experiment or someone asking a mutual friend about a blind date beforehand, intelligent failures still require preparation. A chemistry team isolating RNA strands succeeded only after a lab member dug up an obscure 1960s paper on the reagent glyoxal, an act of homework that turned repeated failed reagents into a working method.

decision-making

Simulators and R&D departments let organizations fail small instead of failing big.

Rather than letting a new pilot's mistakes play out with 200 live passengers, put them in a simulator first. Rather than letting a failed product idea reach customers, test it inside an R&D group. Both isolate the cost of failure from the people who would otherwise bear it.

risk-management

Tolerating repeated failure requires reframing setbacks as expected, not as a verdict on your competence.

Edmondson describes this as a form of wisdom and character: treating a setback as 'that's disappointing, but it's okay' rather than 'I don't like it.' She credits Thomas Edison's line, 'I haven't failed, I've just found 10,000 ways that don't work,' as scientifically valid rather than merely optimistic, because each failed attempt genuinely narrows what remains to be tried.

resilience

A single clerical error can trigger a cascading real-world collapse even when quickly corrected.

In 2009, a British government registrar (Companies House) mistakenly reported that Philip Davison-Sebryi's 134-year-old company, Taylor and Sons, was going out of business, confusing it with an unrelated company, Taylor and Son. Even though the mistake was corrected within days, the reputational damage triggered canceled contracts and a debt spiral that ultimately did put the real company out of business.

failure-taxonomy

Books referenced

Right Kind of Wrong: The Science of Failing Well - Amy Edmondson - Edmondson's book laying out her taxonomy of basic, complex, and intelligent failures, referenced throughout as the source of her framework.

Media referenced

My Unsung Hero - podcast - Hidden Brain's sister show; the closing segment features listener Susan Prescott's story about her English teacher.

Companies

Toyota - Cited as the model for catching errors early via the Andon cord, which lets any assembly-line worker stop production to flag a possible problem.
NASA - The Columbia space shuttle disaster is analyzed as a complex failure where an early warning sign (a foam strike) was dismissed by senior managers.
Companies House - British government business registrar whose clerical typo (confusing 'Taylor and Sons' with 'Taylor and Son') triggered the opening anecdote about a real company's collapse.
Air Florida - Flight 90 crashed into the Potomac River in 1982 after the crew skipped the anti-ice checklist item out of habit, illustrating a basic failure.

Techniques and frameworks

Taxonomy of failure (basic, complex, intelligent) - Edmondson's central framework: basic failures are single, preventable slips; complex failures are multiple small factors lining up (Swiss cheese model); intelligent failures are useful, hypothesis-driven experiments in new territory.
Swiss cheese model of failure - Borrowed from error theorist James Reason; complex failures occur only when multiple independent defects (holes) line up, illustrated via a morphine overdose case with seven contributing factors.
Andon cord - Toyota production-line practice where any worker can pull a cord to flag a possible problem; a team leader investigates immediately, and the line only stops if the problem is confirmed.
Small losses / bounding intelligent failure - Deliberately limiting the size, cost, or duration of an experiment (a drink instead of a weekend visit, a flight simulator instead of a real cockpit) so that failure, if it happens, stays cheap.

Summary

Hidden Brain host Shankar Vedantam sits down with Harvard Business School professor Amy Edmondson to unpack why organizations and individuals get failure so wrong, treating every mistake as equally bad when in fact mistakes come in fundamentally different kinds that call for different responses. Edmondson opens with the origin story of her research: as a young scientist tracking medical errors across hospital teams, she expected better teamwork to correlate with fewer mistakes. Instead her data showed the opposite, better teams appeared to have more errors. The resolution turned out to be a single extra survey question she'd added almost as an afterthought: whether mistakes were held against people. Teams where speaking up felt safe weren't making more mistakes, they were simply willing to report them. That insight, that punitive cultures don't eliminate error but only its visibility, became the foundation of Edmondson's decades of work on psychological safety.

From there the conversation builds out Edmondson's taxonomy of failure. Basic failures are simple, preventable slips, the kind that happen when a task feels so routine that vigilance lapses, as in Edmondson's own sailing accident or the 1982 Air Florida crash where a crew skipped a checklist item "habitually." Complex failures result not from one cause but from several small factors lining up, illustrated through James Reason's Swiss cheese model and applied to both the Columbia space shuttle disaster and a pediatric morphine overdose with seven contributing causes. In both cases, the failure was preventable in principle because multiple independent signals existed, but no single person or system caught the alignment in time.

The most counterintuitive category is intelligent failure: the useful, even necessary, byproduct of genuine exploration. Edmondson lays out three markers of an intelligent failure: it happens in new territory, it's guided by a real hypothesis, and it's deliberately kept small. She illustrates this with a chemistry lab where 90% of hypotheses fail but each failure narrows the path forward, with Thomas Edison's dogged search for light bulb materials, and, memorably, with her own parents' courtship, where her mother's disastrous first blind date led her to agree to a much lower-stakes second introduction (a drink, not a weekend) that turned into a lasting marriage.

Threaded through the whole conversation is a practical argument against blanket policies. Neither "zero tolerance for failure" nor Silicon Valley's "fail fast, fail often" is right in all contexts; the correct posture depends on which kind of failure a given domain is prone to. Operating theaters and airline cockpits need disciplined, checklist-driven execution aimed at eliminating basic failures. R&D labs and personal growth need the freedom to run small, hypothesis-driven experiments. Toyota's Andon cord embodies the middle path: a cheap, fast mechanism for surfacing possible complex failures before they compound, treating the cost of stopping the line as an investment rather than a loss.

The episode closes on a lighter note with the show's regular "My Unsung Hero" segment, in which listener Susan Prescott credits her twelfth-grade English teacher for quietly reframing a stutter-related fear of public speaking into confidence that shaped her career as a corporate trainer.

Notable Quotes

"When you insist that we must have error-free performance or error-free lives, the main thing that happens is not that error goes away, it's that you stop hearing about it." - Amy Edmondson

"Failures? I haven't failed. I've just found 10,000 ways that don't work." - Amy Edmondson, quoting Thomas Edison

"Anytime you hear yourself or someone else saying, oh, I can do this in my sleep, watch out. You can't." - Amy Edmondson

"I'm a fallible human being who overly trusts my prior experience and often fails to be curious enough about, well, maybe this one's different... So it's tragic, but there are no bad guys here." - Amy Edmondson