We spend months designing experiments to detect failure. We spend almost no time imagining it in advance.
That's the premortem: gather the team, say "it's 18 months from now and this completely failed," then work backwards. Known in decision-making circles. Almost never used in research.
Why? Because in academia, questioning a PI's pet idea feels like insubordination. Because sunk costs start on day one. Because "what could go wrong?" sounds like pessimism — but it's not. A premortem is structured. Pessimism is just morale damage.
The output isn't a veto. It's a checklist: "If X happens, we already decided Y." You're not preventing failure. You're preparing for it.
A lightweight version takes 20 minutes. Before a grant submission or new collaboration, ask the room: "It's a year from now. This failed. Why?" Three rounds. That's it.
The molecule is optional. The failure mode is universal. The same premortem works for a PhD thesis chapter, a committee evaluating proposals, or an engineer scoping a feature.
When did you last imagine your project failing — on purpose?