The Rumsfeld matrix has four quadrants. The top two — what we know we know, and what we know we do not know — are the comfortable ones. The bottom two are where the trouble lives. Unknown knowns is the quadrant for tacit knowledge, suppressed findings, and silent assumptions. Unknown unknowns is the quadrant for the genuine surprises — the system-level shocks, the equity dimensions nobody thought to disaggregate by, the second-order consequences in adjacent sectors.
Most programme design lives in the top row. Most evaluation failure lives in the bottom row. This is a note on the meeting that drags the bottom row onto the page before the launch — the pre-mortem.
What a pre-mortem actually is
A pre-mortem is a post-mortem for a project that has not failed yet. The framing is simple. Gather the team in a room. Say: Imagine it is eighteen months from now, and the project has failed. Write the obituary. Set a timer. Give everyone twenty minutes to draft, alone, on paper.
Then read them out, one by one, without arguing.
That is the whole exercise. The structure does most of the work. People who would not raise a concern in a planning meeting will write it down in a failure scenario because the scenario gives them permission. The exercise does not ask whether the project will fail. It asks: assuming it has, what killed it?
Why the obituary is the unit
Three things happen when you ask people to write the obituary rather than list risks.
First, the cause of death is concrete. A risk register reads “stakeholder misalignment, medium probability, medium impact”. An obituary reads “the District Collector transferred in month four, the new one had been briefed against the programme by the previous contractor, and we never recovered the access we lost in that meeting”. The second is a thing you can plan against. The first is a thing you can file.
Second, the obituary forces sequencing. Real failure has a date. A risk list is timeless. An obituary makes the team commit to when the failure begins, and that commitment usually surfaces a window the planning had quietly assumed was safe.
Third, the obituary lowers the social cost of raising a concern. In a planning meeting, naming a risk is naming a colleague’s optimism as wrong. In a failure scenario, naming a risk is helping the team see something nobody is yet defending. The pre-mortem turns a confrontation into a collaboration.
How it fits the matrix
Run on the matrix, the pre-mortem populates the bottom row directly.
Unknown knowns surface in the obituaries that begin “we always knew that…” — the tacit knowledge the team had but had not voiced. The frontline staff who had run this programme before and had not been invited to the design meeting. The findings from the last evaluation that were politically inconvenient. The assumption baked into the choice of district that nobody had checked against the political map.
Unknown unknowns surface in the obituaries that begin “what we did not see coming was…” — the genuine surprises the team had not modelled. The cross-sectoral consequence. The shock in an adjacent system. The population the sample frame could not reach. These are harder to surface, but the pre-mortem still helps: even the act of imagining failure forces the team to scan further out than the planning document allowed.
When to run it
Three good moments.
Before the proposal is finalised. A pre-mortem at the proposal stage costs an afternoon and may prevent the programme being designed around an assumption that fails in month three.
Before the evaluation is commissioned. A pre-mortem on the evaluation itself — imagine the evaluation produced findings nobody used; why? — is the most under-rated meeting in MEL practice. Most evaluations fail at use, not at design, and the use failures are usually pre-mortem-able.
Before scale. A pre-mortem on the scale plan — imagine the pilot worked but the scaled version did not; why? — catches the assumptions that were carried forward from a context that no longer holds.
The artefact
The output of a pre-mortem is not a report. It is a single sheet — a populated bottom row of the Rumsfeld matrix — pinned above the desk for the duration of the project. Each obituary line gets translated into one of two things: a planning change (we will now design against this), or a watchpoint (we cannot prevent this, but we will know if it begins).
The Knowledge Map canvas at /canvas/rumsfeld-matrix is built to hold this output. Print it. Fill in the bottom row in a pre-mortem session. Pin it up. Come back to it in month three.
The cause of death usually has a date on it already. The pre-mortem just lets the team see the date before the funeral.