Pre-Mortem
An exercise in which a team imagines a project has already failed and explains why, surfacing risks that a forward-looking risk assessment misses.
Forward-looking risk assessment produces a polite list. The pre-mortem changes the framing: "it is eighteen months from now and this has failed badly. Write down what happened."
The psychological difference is real and well documented. Asked to predict risks, people hedge; asked to explain a failure that has already happened, they are specific, and they raise concerns they would otherwise have thought impolitic.
The mechanics: individual written accounts first, before any discussion, which prevents anchoring on whoever speaks first and surfaces the true spread of concern. Then share, cluster, and convert the recurring themes into mitigations with owners.
What it reliably surfaces that a risk register does not: organisational risks — the sponsor leaving, the funding being cut, a reorganisation mid-programme; the dependency everyone privately doubts; and the assumption nobody wants to challenge in front of its author.
The related habits worth building for architecture specifically: failure-mode enumeration for each dependency ("what happens when this is slow, not just when it is down" — latency is more revealing than failure); asking what would have to be true for a design to fail catastrophically; and treating every incident as evidence about a class of failure rather than as a single defect to fix.