What goes into an improvement plan that actually works?

Why do so many plans fail quietly?

Because they were never written in a form that could fail loudly. A plan that says "improve receiving" cannot be wrong. A plan that says "lead time through receiving on the day shift, measured weekly, owned by the shift lead, reviewed in ninety days" can be wrong, and that is what makes it a plan.

What are the four parts?

Which constraint?

Name the step. Not the department, not the problem, the step whose capacity bounds the flow and which the map and the queues have confirmed. If the constraint has not been found, the plan is premature.

Which measure?

The one that will show the change. Usually throughput at the constraint or lead time through the affected steps. Never the cycle time of a step that was already fast, and never a count of activities completed. A plan without a measure is a wish, and a plan with the wrong measure is a wish with a chart.

Who owns it?

One person, named, who controls the thing being changed. For a batch-size change that is the shift lead who releases work. For a change to how vehicles are booked in, it is whoever controls the booking. An owner who cannot change the thing is not an owner.

When is it reviewed?

A date, written down at the start. The review compares the measure before and after. It is not a meeting about how hard everyone tried.

Where does the plan come from?

From the people doing the work. They already know where the waiting is. Ask them before proposing anything. The six supervisor interviews in week two of my current review exist for exactly this reason: thirty minutes each, before any proposal is drafted, so that the plan is built from what the floor knows rather than presented to it.

A plan that the floor recognises as its own is one the floor will carry when the consultant has gone.

Why pilot on one shift?

Because a pilot is a measurement, and a measurement needs a comparison. Change one shift, leave the others as they are, and the difference in the named measure tells you whether the change worked. Roll out everywhere at once and you have no comparison, only a before and an after confounded by everything else that changed that month.

One shift is also small enough to undo. If the pilot shows the measure going the wrong way, the plan is corrected on one shift rather than across the whole depot.

What does the review look like?

A single comparison: the named measure before and the named measure after, over the same period, on the same shift. If the measure moved, the change is rolled out and the constraint is checked again, because it may have moved. If the measure did not move, the plan was wrong and that is useful to know early.

Write the review as that comparison. Not a narrative of effort, not a list of actions completed. Those feel like progress and prove nothing.

What is the first action in almost every case?

Measure at the constraint before changing anything. In my study notes this appears as the answer to a scenario question, and it is also what I would say to a client who wants to start fixing things on day one. A week of measurement before the first change is cheaper than a month of changes in the wrong place.

What does a finished plan look like?

For a fixed-fee review it is the third deliverable in the engagement letter, after the current-state map and the prioritised constraint list, presented to the operations director for acceptance. It is short. One page per constraint: the step, the measure, the owner, the pilot shift and the review date. Anything longer is hiding a missing measure.

Q&A

The questions clients ask most about improvement plans are answered below.

Last updated 1 October 2026.