The Strange Power of Imagining Failure

By Nathan Waterhouse · 2026-05-27

"I want you to imagine it's a year from now, and your current strategy has failed."

As the words left my mouth, I watched the faces in the room settle into that same flicker of trepidation I always feel when I'm about to ask a group of leaders to do something slightly mad. I brace for it every time.

But as soon as I say go, something different happens. People don't go awkward, the way they can when I ask them to draw something. They turn to face each other, and they get straight to work. There's something strangely energising about imagining failure.

The room begins to get very specific. Someone names a particular stakeholder who quietly disengaged. Someone else describes how the new framework got translated, on its way through the operations function, into something more like the old framework with new vocabulary. A regional lead points out that two months in, the people who were meant to champion the change had defaulted back to their default behaviours. Far from feeling bleak, the conversation becomes clinical, granular, almost relieved. The room is no longer hypothesising about possible risks. It's narrating an event it can see in its mind.

I've used the premortem in commercial strategy programmes, change initiatives, transformation work. It lands well in almost every room I've taken it into. What I've come to notice is that people don't experience it as an exercise in pessimism, even though that's how it looks on the agenda. They experience it as an exercise in clarity. They walk out energised by what they've surfaced, not deflated by it.

For a long time I didn't understand why. The neuroscience, it turns out, has a good answer.

What the 1989 study actually found

The premortem owes its popularity to Gary Klein, the psychologist who named the technique in a 2007 Harvard Business Review piece. Klein drew on a much older piece of research by Deborah Mitchell at Wharton, Jay Russo at Cornell, and Nancy Pennington at the University of Colorado, published in 1989 in the Journal of Behavioral Decision Making. The paper was titled, rather wonderfully, Back to the Future: Temporal Perspective in the Explanation of Events.

The headline finding is the one most people quote. Imagining that an event has already occurred, as opposed to imagining that it might occur, increases the ability to correctly identify reasons for future outcomes by about thirty per cent. That figure has been doing a lot of work in management writing ever since.

What's less often quoted is the buried finding. When you ask people to imagine an event as having already happened, they don't just generate more reasons. They generate a different kind of reason. Mitchell, Russo and Pennington noted that explanations for "sure" events were longer, were expressed in past tense, and contained a much higher proportion of what they called episodic reasons. These are reasons that read like remembered scenes rather than general principles. Specific people doing specific things in specific moments. Not "stakeholder resistance" but "Anna in regulatory pulled the team off it in week three". Not "communication gaps" but "the August handover was thin because two key people were on leave."

The 30 per cent isn't a trick of phrasing; it's the byproduct of the brain being asked to do something it's surprisingly good at.

Why the brain finds the future easier when it's already happened

The reason the premortem works lies in a finding from cognitive neuroscience that took most of the 2000s to settle, and which still surprises me when I think about it. Imagining the future and remembering the past use the same machinery.

Daniel Schacter and Donna Rose Addis, then both at Harvard, proposed what they called the constructive episodic simulation hypothesis. Their argument, supported by fMRI evidence and patient studies, was that the same "core network" of brain regions (the hippocampus, the ventromedial prefrontal cortex, the medial parietal cortex) activates whether you're recalling a holiday from five years ago or imagining a dinner you might have next month. The brain doesn't really have a separate "future" system. It has a memory system that can be run forward as well as backward.

Patients with damage to the hippocampus lose access to detailed autobiographical memory. They also, strikingly, lose the ability to imagine specific future events in any vivid detail. The two capacities turn out to be the same capacity. If anything, studies of episodic future thinking suggest the hippocampus works harder imagining a future scene than recalling a past one, because it has more to construct from scratch.

This is, I think, why the premortem feels so different from a normal risk discussion in the room. Asking "what could go wrong?" engages a more abstract reasoning mode, where people generate categories of risk from general knowledge. Asking "imagine this has already failed, why?" recruits the autobiographical memory system. The brain reaches not for general principles but for the kind of specific scenes it uses for remembering. People stop reasoning hypothetically and start, in a real sense, remembering a future that hasn't happened yet.

Why the room has to be safe

There's a condition I've skated over, and a recent conversation between Adam Grant and Brené Brown on their podcast brought it back into focus. They were debating what actually makes a premortem work. One view holds it's purely about psychological safety: everyone in the room already knows the risks, and the only question is whether they feel safe enough to say them aloud. Grant disagreed. A good premortem, he argued, doesn't just give permission to voice what's unsaid. It generates genuinely new information, the kind that takes anticipatory thinking and systems thinking, skills most teams haven't built. And you can't build those without safety, so you end up doing both at once.

That matched something I've long believed about facilitation. The episodic machinery I described only fires if people feel able to narrate the real ways things might go wrong, including the ones that implicate them personally. Psychological safety, the concept Amy Edmondson put into the world, is the precondition for the whole exercise, which is why I put so much effort into making a room feel safe before I ever ask anyone to imagine failure. A premortem in a guarded room produces tidy, defensible risks. A premortem in a safe one produces what people actually fear.

The connection to the room

I wrote recently about how the shape of a question determines what becomes thinkable in a room. The premortem is the cleanest example I know of that principle. What could go wrong? and Imagine this has failed; why did it fail? ask about the same future, but they reach for different systems, and the second one pulls out the risks that actually sink a strategy: the person who quietly stops championing it, the meeting that didn't happen, the budget cycle that fell on the wrong side of approval. The specific, episodic ones a risk register never surfaces.

The premortem doesn't ask people to predict the future. It asks them to remember it. And a room that can remember its own failure, in detail, before it happens, is a room with a real chance of avoiding it.

If you run one, watch what happens in the first few minutes. The shift from "risks we should list" to "things we can already see happening" is usually visible in the room. I'd be curious to hear from anyone who's used the premortem in a setting where it surprised them, or where it didn't quite land the way they expected.