What boards should ask in month three

Month three is the earliest honest checkpoint and the last cheap one. It is early enough that re-sequencing costs weeks rather than write-offs, and late enough that the programme has produced evidence. In the programmes we audit, the ones that failed at month twenty had almost without exception already failed the month-three questions. Nobody had asked them.

The activity report answers the wrong question. It says whether the programme is busy. The four questions below say whether it is loaded, owned, moving and measured, which are the properties that predict whether it will hold.

First: how many priorities is the organisation actually carrying? Count everything in flight, not only the programme’s three. If the honest register says nine, the programme’s three are fiction. Second: for each commitment, name the single owner and the frozen baseline. Hesitation on either is the earliest failure signal we know. Third: what decisions has the programme needed, and how long did each wait? Latency at month three compounds, and a programme queuing three weeks for decisions now will queue six by month twelve.

Fourth: what has the middle layer stopped doing to make room? If the answer is nothing, the capacity for the change is being borrowed from the operation, and the operation will take it back, usually around month nine, as slippage nobody can quite explain.

A board that asks these four questions at month three buys itself options that do not exist at month twelve. It can stop something cheaply, re-sequence while the phases are still fluid, or hold course on evidence rather than optimism, and none of those options survives to the review a board would normally wait for, by which point the cheapest correction available is usually an apology. The programmes in the top quartile of our cohort were not the ones that never drifted. They were the ones whose drift was caught while correcting it was still a Tuesday-afternoon decision.

None of the four questions requires programme expertise. They require a board willing to be unsatisfied by green. The register, the owner, the queue and the stop list: a transformation that can answer all four in writing is, in our data, already most of the way to being one that holds. We should be plain that this is association rather than proof, because we have never run the experiment in which a board asks the questions and then ignores the answers.