From strategy to implementation: closing the delivery gap

A strategy can be analytically sound and still not occur. The document is approved, a launch is held, and a year later the operating work resembles what preceded it. The usual verdict is weak implementation, which treats implementation as a disposition: effort, ownership, political will. Disposition matters. It does not explain the cases in which capable institutions adopt a plan and then continue, quite rationally, to run the system they already have. The plan asked them to want a different outcome. It did not give them a different way of working.

The gap is specific, and it can be located in the document. Most strategies specify an outcome and a list of activities. They do not specify the operating arrangements required to carry those activities: which office holds the mandate, which routine changes in an ordinary week, which decision rights that office already has, which information would tell it the work is off course, and what consequence follows if the work is not done. A results chain that stops at activities leaves the assumptions unstated. An activity such as “training will be conducted” is not an assumption about change. The assumptions that matter concern authority, money, time and information. If those are false, the activities can be reported as complete while the outcome does not move, and the monitoring system will record success at the wrong level.

Approval is therefore the wrong moment to begin this design. By the time a strategy is tabled for endorsement, the institution that must carry it, the budget structure that must fund the routine, and the information that must show whether the routine is happening ought already to have been tested against the way the organisation actually works. A plan that reaches approval without that test is not an early draft of implementation. It is a document whose hardest choices have been postponed. Postponement, more than a lack of effort after the launch, is how a sound objective fails.

Public and development programmes are not delivered by the strategy. They are delivered by organisations that already have mandates, an establishment, budget lines and habits. Those organisations are not a blank site on which a new operating model can be drawn. A strategy that depends on a parallel structure, a new unit with no authority over the budget or the staff who do the work, or a task force that meets and then disperses, has not engaged the institution. It has stepped around it. Parallel structures can serve a time-limited design task. They cannot be the lasting home of a function the permanent organisation was never asked to absorb, and they collapse, predictably, when the external support that convened them ends.

For each major commitment, the question is who already has the mandate. If the answer is no one, the strategy has created a task without an owner. Coordination language will not supply the legal or administrative authority the task requires. If the answer is an office that is already saturated — its establishment filled, its calendar taken by statutory returns, its operational budget committed — the strategy has added work without removing any. Both failures are visible before approval, if the establishment and the budget are examined rather than the organogram drawn for the strategy annex. An organogram shows where a box was placed. It does not show whether that box can compel a budget release, a posting or a change in clinical or operational practice.

Implementation support is the work of fitting the commitment to that owner. It asks which staff they have, which money they control rather than which money is described as available to the sector, which decisions they can take without a fresh approval, and which constraints they cannot ignore: a procurement rule, a posting cycle, a commodity pipeline, a statutory report that already consumes the same people. A design that assumes those constraints away will look clean in the document. It will be renegotiated, informally and incompletely, by the people who have to carry the week. That renegotiation is then recorded as an implementation gap, when it was a design gap that was refused at the drafting stage.

Activities in a plan are not routines. An activity can be launched, counted and reported once. A routine is what happens on an ordinary week when no one is visiting: the meeting that is actually held, the list that is actually reviewed, the threshold that triggers a response, and the person who can reallocate effort without waiting for a new instruction from the centre. Strategies stall in the space between the two. The plan says that supervision will be strengthened, or that data will be used for decision-making. Neither sentence describes a week, so neither can be managed, supervised or found to have failed in time to correct it.

The routine has to be specified with the objective, not after a review has found that nothing moved. For each result worth having, five things should be writable. What will be done differently in the ordinary week, as distinct from what will be launched. Who checks, and on what cadence. What information they will use, from which source, and what that source cannot show. What they are allowed to change — staff time, a commodity request, a visit schedule — without seeking fresh approval. And what is escalated, to whom, and after how long. If any of these cannot be stated, the strategy is not ready to implement. Further drafting of objectives will not repair it. The missing piece is a decision about ownership and discretion, which is harder than a decision about wording because it assigns a responsibility someone will later have to meet.

The same specification is where work should be retired. A new routine that does not say what the office will stop doing is an addition to a saturated week. It will be the first thing dropped when the week is full, and the strategy will record the drop as a failure of commitment. Choosing what stops is part of implementation design. Leaving it unstated transfers the choice to the officer who is already overcommitted, and then blames that officer for the result.

Accountability in a strategy is often a table of names against activities. A name is not a consequence. Accountability is real when a missed action has an effect the institution will actually apply, using authority it already has. That effect might be a delayed release, a supervision visit triggered by a threshold rather than by a courtesy calendar, a report back to the body that approved the plan, or a change in who is trusted with the next assignment. If none of these can be named, the accountability framework is a distribution list. It will be updated, and it will not govern anything.

The consequence has to be legitimate. An advisory team cannot invent a sanction, a budget control or a posting power the institution does not possess, and it should not write as though a steering committee held authority that the financial regulations do not give it. The useful work is to find the controls that already exist and attach them to the routine: which report already goes to an accounting officer or a board, which release is already conditional, which supervision function already has the right to visit. It is also a legitimate conclusion that a plan is not implementable, in the plain sense that no one can say what happens if delivery slips. That conclusion is more useful than a matrix implying consequences no one will apply.

Where several agencies share the work, failure usually sits at the handover rather than inside either agency’s own plan. One office treats its part as discharged when a commodity is dispatched, a trainee is certified, or a transfer is approved. The next office receives a delay, a mismatch or a missing input, and has no signal that was agreed in advance. The design should name the interface. What one agency owes the next, by when. What “done” means in a form the receiver can check. How a delay becomes visible before the quarter is lost. Programmes do not stall only for lack of effort inside institutions. They stall in the space between them, where no one’s routine includes the handoff, and where each agency can report its own activity as complete.

The passage from an approved plan to operating work is where many advisory engagements prefer to stop. The analysis has been accepted, the document has a sponsor, and the difficulties that remain are classified as matters for the client to resolve in implementation. Some of them are. Many were visible while the plan was being written, and were left out because they were awkward: a mandate that does not cover the task, a data source that cannot support the indicator the strategy intends to publish, a budget line that cannot pay for the routine, an office whose week is already full. Meeting these during implementation is not a surprise. It is the cost of having treated knowable constraints as details to be handled later.

A strategy closes the delivery gap only when those constraints are part of the design. The results chain includes the institutional assumptions, not only the activities. The routine is specified at the level of a week. The information is drawn from a source that can bear the inference, with the limits of that source stated. The consequence of slippage is one the institution will apply with authority it already has. None of this requires a longer document. It requires that the drafting process be willing to stop when a commitment cannot be attached to an owner, a routine, a source and a consequence. A commitment that cannot be so attached is not yet a strategy. It is an intention.

The test after approval follows from that design. Within a cycle, often within six months, a manager should be able to point to the routine that changed, the information used to watch it, the person who can adjust it within their existing authority, and something in the week’s work that is different because the plan exists. If they cannot, the institution does not yet have a strategy in the operating sense. It has a document that was approved. Approval was never the scarce step. The scarce step was specifying, before the launch, how the institution would carry the work on an ordinary week, and what would be true if it did not.