Make Your Decision Twice: Once Forward, Once From the Future

Ask a team “what could go wrong with this plan?” and you get a short, polite list. Vague risks, hedged language, nobody wanting to be the pessimist in the room.

Ask the same team to imagine it’s one year from now and the plan was a total disaster — then ask them to explain why — and something strange happens. The list gets longer. The risks get specific. And the person who was quiet in the first round suddenly has three things to say.

Same team, same plan, same amount of time. The only thing that changed is the tense.

The trick is grammatical, not analytical

This is the premortem, a technique the psychologist Gary Klein popularized in the 1990s, and it rests on something researchers call prospective hindsight — explaining a future event as though it had already happened, rather than speculating about whether it might.

The original 1989 study behind this found something counterintuitive: people are dramatically better at explaining an event they’re told is certain than one they’re told is merely possible. Ask “why might this fail?” and people hedge, generalize, undersell. Ask “this failed — why?” and the same people produce roughly 30% more reasons, most of them concrete and actionable rather than vague.

Nothing about the plan changed. Only the grammar did — from conditional to past tense. Human reasoning, it turns out, is much better at explaining than at predicting.

Why “what could go wrong” doesn’t work

The conventional risk review has two failure modes, and they compound each other.

Cognitively, “what could go wrong” asks people to search an open-ended space of possibilities, which is hard. “Why did this go wrong” asks people to explain a single, fixed outcome, which is easy — it’s the difference between free-associating and answering a specific question.

Socially, raising doubts about a plan the team already committed to is an act of dissent. It marks you as the negative one, the person who doesn’t believe. Daniel Kahneman called the premortem one of the most useful de-biasing tools precisely because it flips this dynamic: once failure is the assumed premise, contributing reasons for it isn’t pessimism — it’s the assignment. The socially risky move becomes the expected one.

The protocol

The mechanics are simple enough to run in under an hour:

1. State the plan. Everyone starts from the same understanding of what’s being decided.
2. Set the premise. “It’s one year from now. This failed completely.” Not might fail — failed.
3. Write independently. Each person silently lists every reason they can think of, before anyone talks. This matters — round-robin sharing after independent writing surfaces more and more varied reasons than open discussion, where the first opinion anchors everyone else.
4. Share round-robin. One reason per person per turn, no debate yet, all of it recorded.
5. Cluster and rank. Group similar failure modes, then rank by likelihood and impact.
6. Revise the plan. Build safeguards, contingencies, or go/no-go checkpoints against the top-ranked failure modes.

The output isn’t a feeling of having been thorough. It’s a specific list of the ways this particular plan is fragile, ranked by how much it should worry you.

The forward pass isn’t enough on its own

Here’s the part that matters for how you actually use this: a premortem isn’t a replacement for normal decision analysis. It’s a second pass, run from a different vantage point.

Forward pass — the decision-making you already do: define the goal, weigh the options, run the numbers, model the uncertainty, pick a path.

Backward failure pass — the premortem: assume that path failed, and explain why, in detail.

You can extend this a third way with backcasting: instead of assuming failure, assume success — a specific, vivid future where the decision worked out exactly as hoped — and work backward from there to identify what had to be true along the way. Comparing the backward-from-failure list against the backward-from-success list often reveals the same handful of variables driving both outcomes: the things that, if they go well, mean success, and if they go badly, mean disaster. Those are the variables worth watching closely, not the ones that only appear in one list.

A plan only earns commitment once it survives all three lenses — forward, backward-failed, backward-succeeded — not just the first one.

Where this earns its keep

Not every decision needs three passes. A premortem on what to have for lunch is a waste of everyone’s time. The technique pays off specifically when:

• The decision is expensive to reverse — an acquisition, a launch, a lease, a hire you can’t easily undo
• The uncertainty is real, not just unmodeled — no amount of spreadsheet precision substitutes for imagination when the future genuinely doesn’t resemble the past
• The room has a bias toward agreement — a founder’s pet project, a plan senior leadership already loves, anywhere dissent feels costly

For routine, low-stakes, reversible choices, skip it. The value of a premortem is proportional to how much a bad surprise would cost you.

Running it on your own decision

You don’t need a team to run a premortem — you need fifteen minutes and the willingness to write down uncomfortable sentences. Before you commit to the job, the move, the investment: write “it’s one year from now, and this was a mistake,” and then finish the paragraph. Don’t soften it. Let the failure be specific.

That list is exactly the kind of input a Monte Carlo model wants — each concrete failure mode becomes a variable with a probability and a cost, instead of a vague worry sitting in the back of your mind. Run the numbers forward, run the failure story backward, and see if they agree. When they don’t — when the spreadsheet says “proceed” but the premortem surfaces three specific, plausible ways this goes badly — that disagreement is the most valuable thing either exercise produced.