Test Each Finding Against the Change Already Funded

The kickoff that opened with a warning

A plant leader opened a project kickoff by telling me the timing was bad. A capital expansion had just cleared his board, equipment for part of it was already quoted, and some of it was already on order. A major product was moving to a sister plant within weeks, which would turn a dedicated crew into a flex team running a different line. He was already pulling people off one line most days and holding back on backfilling, because he knew the volume shift was coming at him.

Then he said the part worth writing down. Over the next several weeks we might discover something real on the floor, and he did not know how actionable it would be, because many of the ideas we would bring forward were already consumed by the expansion.

That is not resistance to the work. That is a leader telling you his plant has a change date and your findings do not.

A finding has a configuration, not just a number

Every observation on a floor is a measurement of one specific arrangement: this layout, this product mix, this crew structure, this sequence of steps. The number is the visible part. The arrangement that produced it is the part that expires.

When a funded project or a volume transfer lands, open findings do not all fail the same way. They separate into three fates, and the difference decides who may claim the value.

A finding is consumed when an approved project already captures it. The handling position you would recommend removing disappears when the new equipment goes in, and that value belongs to the capital case that paid for it. Recommending it again does not create a second saving; it creates a second claim on one saving.

A finding expires when the change removes the conditions it depended on. A crew balance measured against a product that is leaving, a sequence built around a step that is being relocated, a staffing level sized for a mix that is about to shift. The mechanism you identified may have been real on the day you measured it and irrelevant the following quarter.

A finding survives when it does not touch the changing part of the operation. A method change inside a step that is staying put, a position made unnecessary by a rearrangement the project never reaches, a decision rule about how work is released. These are the findings worth acting on now, and they are usually the smaller ones.

The quiet fourth case is the one that causes arguments. Sometimes the mechanism survives and the magnitude does not. The waiting you measured between two steps will still exist after the change, but not in the same quantity, so the effect is real and the number attached to it is no longer supportable.

Sort the findings before you sell them

Start by getting the change onto paper with dates. Which projects are funded, what physically moves, when each piece goes live, and which products change plants or formats. A leader living through it can usually give you this in ten minutes, and it is the most valuable input in the engagement.

Then assign every open finding a fate and a reason. Survives, consumed, or expires on a stated date. Write the configuration each one assumes next to it. A finding without a stated configuration cannot be defended later, because nobody will remember which arrangement it described.

Attribute value to the change that causes it. When the plant has already decided and paid for something, document it as captured by the plant, not as a test still to be passed and not as a recommendation. This costs you credit in the short term and protects the entire case in the long term, because the first time a claimed saving turns out to belong to a funded project, every other number in the document becomes suspect.

Capture the baseline before the change, even for work you plan to defer. After the mix moves, no comparable before exists, and a saving nobody can measure is a saving nobody will approve. The measurement is cheap while the current arrangement is still running and impossible afterward.

Sequence the tests that remain inside stable windows. A staged validation that steps one variable at a time through the weeks before a changeover will produce a defensible result. A single before and after spanning the change will not, because the change itself explains the difference.

And defer honestly. Work whose value depends on the post change configuration should be scheduled to start after the operation stabilizes, with the condition that restarts it written down. Deferred means scheduled against a condition. It does not mean dropped.

What a well run improvement plan reads like

Every open finding on the list names the layout, mix, and crew it was measured under. Each one carries a fate and, where it expires, a date. Items already funded by a capital project appear in a separate section attributed to that project, with no dollar value assigned to the improvement work. Every finding held for later carries a baseline captured before the change and a written condition that triggers the restart. The savings total on the front page reconciles to the sum of the items marked survives, and nothing else.

Before the next review, take the open finding list and put a change date beside each line. The ones that cannot take a date are the ones to act on this month.