Independent delivery assurance, separated from delivery itself. In the cohort data that split is worth a 37-point gap in value retention, the largest single effect we have measured. Sponsors who commission independent oversight may also be the sponsors who were always going to ask the harder questions, and that is a selection effect we cannot remove from the cohort. We think the separation does most of the work. We cannot prove the sponsor does none of it.
The programme reports green while the operation reports strain, and the board cannot tell which is lying.
The integrator marks its own homework; scope grows and accountability diffuses.
A transformation is stalled at month 14 and the honest state of it is unknown.
Delivered value keeps decaying after the programme team stands down.
The month-0 position, frozen and evidenced before delivery begins. Where it was never captured, we reconstruct it from whatever the organisation kept and mark plainly which parts are reconstruction rather than record, because a baseline nobody trusts is not a baseline.
Outcomes against baseline, load against capacity, decisions against latency. The readout goes to the sponsor rather than the programme — the programme sees the same document at the same time the sponsor does, which is not the same thing as seeing it first.
Where the load test fails, we write down what stops, what holds and what each of those choices costs, in a document that goes to the sponsor whether or not the programme director agrees with it. It is priced rather than diplomatic. That is the part of the job clients like least.
Every outcome tested under MKM-F-003 before the programme is allowed to end. We stay until it passes.
The month-0 position is audited and frozen. The value case is restated in terms that can actually be verified.
A weekly readout of outcomes against baseline, independent of the delivery team’s reporting line.
Failing commitments are surfaced while correction is still cheap. Re-sequencing is recommended in writing.
Outcomes pass the standard or return as open work. Delivered means tested.