Before the decision lands¶

Two practices that belong before commitment rather than after, and that feed each other. Writing the failure account in advance. Naming the conditions that would shift the judgement.

Pre-mortem¶

Before a decision is taken, write a failure account as if it has already gone wrong. Not as a thought experiment. As a document: a paragraph, a short page, a rough memo. Eighteen months from now, the approach has failed. What does the review commission find. What does the independent investigator write. What does the leaked internal report say.

The point is not to predict failure. It is to change what is visible before the decision is made.

Shape of institutional failure¶

Failure tends toward particular shapes worth writing to specifically.

An oversight review that finds: “The advice provided to the governing body correctly identified the primary risk. It did not identify the secondary condition that, in the event, determined the outcome. Decision-makers were not informed of the limits of the assessment.”

An investigator who finds: “The institutional position was presented to the oversight authority with more confidence than the underlying evidence supported. When the evidence changed, there was no mechanism for revising the position.”

A leaked report that says: “The coalition of support assumed for the preferred option did not hold, because the consent of a key stakeholder was taken for granted. The advisor indicated this risk in a preliminary conversation that was not reflected in the formal brief.”

These are not generic failure modes. These are the specific shapes that institutional failure can take: a gap between the formal brief and the informal conversation; a confidence tier that drifted above the evidence; an assumption about a coalition that was load-bearing but unmarked. Writing a failure account toward these shapes surfaces the specific risks in the specific institutional context.

Optimism at commitment¶

At the moment of commitment, optimism is structural. Principals have agreed. Political capital has been spent. The institution is organised around the decision proceeding. Imagining failure requires imagining that collective judgement was wrong. The social conditions do not support it.

Pre-mortem works because it is conducted before that point. A failure account is not a challenge to a decision already made. It is an input to one that has not been. The exercise is legitimate rather than adversarial, which is what makes it possible to do honestly.

Surfacing the assumptions¶

A pre-mortem surfaces the assumptions the brief treats as given. That a coalition of support required for implementation will hold. That a threat actor will behave as modelled. That an institutional stakeholder whose consent is assumed has actually given it. That technical preconditions are in place. That the timeline is achievable.

These appear obvious in retrospect. A pre-mortem makes them visible before the fact, when they can still be examined or at least acknowledged as risks.

A short failure account that names the load-bearing assumptions and writes a scenario in which they fail is more useful than a comprehensive risk register that treats all risks as equal.

Indicators of revision¶

Every assessment carries an implicit question: what would change this judgement. Indicators of revision are the answer, named explicitly before the brief leaves the room.

Two or three observable conditions that would shift the analysis, not hypothetically but specifically.

“If the technical baseline review in Q3 shows X, the risk rating moves.”

“If fewer than two-thirds of the relevant organisations participate in the pilot, the stated objective cannot be met and the recommendation changes.”

“If no significant activity is observed in the next sixty days, the threat assessment is revised downward.”

Named in advance, these do two things. They are more honest than waiting for surprise to revise a judgement, because the conditions for revision are set before any pressure exists to hold the original position. And they give the brief a second life: the original assessment becomes a tracking document, revisable against the named conditions rather than simply ageing out of relevance.

Positions once agreed often generate their own inertia. A position assessed as accurate at the time of agreement tends to get carried forward regardless of whether the indicators that would have changed it have since materialised. Naming the indicators in the brief is not a guarantee that the institution will act on them. It is a record that the advisor identified them.

Composing pre-mortem and revision¶

A pre-mortem surfaces the failure modes. Indicators of revision are the practical output: conditions under which the judgement changes, named while there is still time to name them without pressure.

A brief that has been through a pre-mortem and carries named indicators of revision is not a more pessimistic brief. It is more honest, and more durable. Failure modes are visible. Revision conditions are stated. The decision-maker is not flying on either side of the commitment without instruments.

A security institution operating under a formal mandate is considering whether to endorse an information-sharing arrangement between a subset of its designated partner organisations. The advisor’s recommendation is to support the arrangement, subject to governance conditions in the annex.

The pre-mortem. Eighteen months from now, the oversight review finds:

“The arrangement was endorsed without adequate consideration of the technical capacity differential between participating organisations. Partners with lower capacity were effectively excluded from meaningful participation, which undermined the stated objective of improved collective situational awareness. Governance conditions in the annex were not sufficient to address this differential. The advisor’s recommendation did not flag the capacity differential as a material risk.”

Working through that paragraph: the load-bearing assumptions were that all participating organisations had adequate technical capacity and that governance conditions were sufficient to handle differential capacity. Neither was examined in the brief. The pre-mortem surfaces both before the endorsement is given rather than after the review has found them.

The indicators of revision that follow from this exercise:

One: if the capacity self-assessment submitted by participating organisations in the first ninety days shows a differential above a defined threshold, governance conditions are revisited before full implementation.

Two: if fewer than two-thirds of eligible organisations participate in the pilot phase, the stated objective of collective situational awareness cannot be met by this arrangement and the recommendation is revised.

Three: if the oversight function flags compliance concerns in its six-month review, the endorsement is suspended pending investigation.

These are now in the brief. The decision-maker can take the recommendation having seen the conditions under which the recommendation changes. The advisor has not predicted failure. They have been honest about what failure would look like and what would signal it arriving.