The three types of failure — and why improv is the best place to practise the right kind

Not all failure is the same

Most organisations have one setting for failure: avoid it. A mistake is a mistake, and the instinct is to find who's responsible and make sure it doesn't happen again. It feels rigorous. It quietly does a great deal of harm.

Treat every failure as a fault, and two things follow. People hide the failures they can — so you lose the chance to learn from them — and they stop taking the risks that progress depends on, because the downside of being wrong is too high. You end up with a team that looks careful and learns slowly.

Amy Edmondson, the Harvard professor best known for the concept of psychological safety, offers a sharper way to think about it. In her book Right Kind of Wrong, she argues that failures come in three distinct types, and the job of a leader is to tell them apart.

Edmondson's three types of failure

Basic failures. These are preventable mistakes in familiar territory — the known process wasn't followed, or attention slipped. Sending the file to the wrong client, missing a step in a checklist, a typo in the figures. There's usually a single, clear cause, and the lesson is clear: tighten the process and the attention. These you want to reduce.

Complex failures. These happen in familiar settings too, and they have several causes that line up at once — a perfect storm. A project derails because a supplier slipped, a key person was off sick, and a signal got missed, all in the same fortnight. No single mistake caused it; a combination did. The lesson is to catch the warning signs earlier and build in more slack. These you also want to reduce, and they're harder to prevent outright.

Intelligent failures. These are the good ones. An intelligent failure is a thoughtful step into new territory, where there was no way to know the outcome without trying. Edmondson sets four conditions: it happens in new, unfamiliar territory, it's a real opportunity to move toward a goal, it's informed by what you already know — so it's a smart bet rather than a wild one — and it's kept as small as possible while still teaching you something. The failed pilot, the experiment that didn't pan out, the new approach that taught you why the old one held. Without these, nothing new ever gets discovered.

The goal is the right kind of wrong

The point of the taxonomy is that these three need opposite responses. Basic and complex failures are worth working hard to prevent. Intelligent failures are worth actively encouraging, because they're how an organisation learns what it couldn't have known any other way.

A leader's real job, then, is to sort failures rather than punish them wholesale. When something goes wrong, the useful first question is which type this was. A basic failure points to a process fix. A complex one points to earlier warning signs and more resilience. An intelligent one points to a lesson worth sharing widely — and a person worth thanking, not blaming.

This is where most organisations come unstuck. They run a single blame response across all three. The result is predictable: people stop reporting basic and complex failures, so those repeat in the dark, and they stop attempting the intelligent ones, so innovation dries up. The organisation manages to get the worst of every category.

Why this rests on psychological safety

There's a reason the same researcher gave us both the failure taxonomy and the idea of psychological safety. They're two halves of one argument.

Sorting failures accurately only works if people will tell you the truth about them — and people only do that when it's safe. In a team where admitting a mistake invites humiliation, every failure gets dressed up, minimised or hidden, and the leader never gets a clean enough picture to sort anything. Psychological safety is what turns failure from a threat to be concealed into information the team can use. Without it, the cleverest taxonomy in the world stays on the page.

So the practical question becomes: how do you build a team where it's truly safe to be wrong, and where people have the reflex to take a smart risk and recover well? You can't lecture that into being. People learn it by doing it.

Why improv is the best place to practise failing well

Applied Improv is, in effect, a controlled laboratory for intelligent failure — which is why "Embrace Failure" is one of our five guiding principles.

Every improv exercise is new territory. There's no script, so people are constantly making informed bets about what to try next, watching them not quite work, adjusting and going again. That's the exact shape of an intelligent failure: a thoughtful attempt into the unknown, kept small, mined for what it teaches. People run dozens of these in a single afternoon — far more reps at failing well than working life usually allows.

And because the stakes are low, the real learning is emotional. People feel the flush of a misstep and discover that the room doesn't collapse — a colleague builds on it, the situation moves on, and the failure turns out to be useful. That experience, repeated, rewires the reflex. Failure stops registering as threat and starts registering as data. The amygdala stands down. People get bolder and more honest at the same time.

It also builds the precise thing Edmondson says the taxonomy depends on: psychological safety. You can't make people feel safe by announcing a policy. You make them feel it by putting them through shared experiences where being wrong is normal, supported and even productive — until they carry that norm back to the meeting room. Improv does this directly, which is why teams describe the change as something they felt rather than something they were told.

Worth being clear about one thing: this isn't a licence for sloppiness. Failing well is a discipline. It means taking smart, bounded risks in service of a goal, recovering quickly, and learning out loud — the opposite of careless. Improv trains that discipline. People leave more willing to attempt the intelligent failure and more honest about the basic one, which is exactly the combination a learning organisation runs on.

Everyone says they want a culture that learns from failure. The teams that actually have one have practised being wrong together, safely, until it stopped being frightening. That's the work — and it's work you can do in an afternoon, and build on from there.