A Savings Baseline Is An Event, Not A Number

The claim that arrives after the result

There is a sequence I have now seen enough times to plan around. A labor study is proposed on a line. Plant leadership objects, firmly and in writing, that there is no opportunity here, that the line is as efficient as it can be, and that the study is a waste of everyone's time. The objection escalates. The study proceeds anyway. Positions come off the line and the line keeps running.

Then the position changes. Not to agreement, which would be easy to handle, but to this: we had already planned that. Anything you are planning, we have planned too.

That claim is almost impossible to refute after the fact, and it is worth understanding why. Not because the people making it are dishonest. Because the only evidence that could settle it was available for about a week, at the start, and nobody wrote it down.

Why "we already planned that" is predictable

Every improvement makes a counterfactual claim. It says the system would have continued in state A and instead reached state B because of a specific set of changes. State B is observable; everyone can count the crew on the line today. State A is gone. It existed, it was measurable, and it stopped existing the moment the first change was made.

So an argument about credit is really an argument about a state that no longer exists, which means it is decided by records rather than by facts. Whoever has the better record wins, and if nobody has a record the argument is decided by whoever has more standing in the room. This is why the credit dispute almost always lands in favor of the incumbent organization. Not through politics. Through evidence asymmetry.

There is a second reason the claim is predictable. Resistance and credit seeking come from the same place. A leader who said there was no opportunity has a real exposure once opportunity is found, and the cheapest resolution available to them is to absorb the result as something already underway. Expect it. A plan that treats it as bad faith will handle it badly; a plan that treats it as a foreseeable response will simply remove the ambiguity it feeds on.

Fixing the baseline as an event

The fix is not better analysis. It is treating the baseline as an event with a date, rather than a number to be reconstructed later.

My rule now is that no labor work starts until three things exist. First, a baseline day on the calendar, agreed by both sides, which is the day the measured starting state is taken. Not a week, not a period, a day. Second, the starting state written at the level the work will change it: headcount by station on that line, on that day, with the shift pattern and the rate it was running. A single plant wide number is not a baseline, because the argument will happen at the station level. Third, an explicit statement of what the organization says it already has planned, recorded before anything moves.

That third item is the one people skip, and it is the most valuable. Ask leadership directly, at the start, to write down every reduction they already have planned for this line with dates. Whatever they list is theirs. Everything else is the project's. The question is easy to answer honestly before the work starts and impossible to answer honestly afterward, which is exactly why it has to be asked first.

From there the discipline is unglamorous. Record every change requested and the date it was accepted, including the changes the plant adopts on its own in response to the work, because those are results and they are the ones most likely to be reclassified later. Hold the baseline day as the start of measurement even when the crew is physically redeployed rather than removed; a position moved off a line is a position the line proved it did not need, and the record should say so in those terms. If it is logged as an attendance or coverage adjustment it will not be countable as anything in six months.

What a well run improvement record looks like

Anyone can name the baseline date for any active improvement and open the document that holds it. The starting state is recorded by station, with crew count, shift pattern, and rate. The organization's own pre existing plan is on file, dated earlier than the first change. Each change has a request date and an acceptance date, and redeployments are recorded as capacity findings rather than as staffing noise. The credit rule was agreed before work began, so the conversation at the end is arithmetic rather than argument.

The next test

Take your largest improvement currently running and try to answer one question from documents alone, without asking anyone: what exactly was the state of this line the day before we started? If you cannot answer it today, you will not be able to answer it when the number is contested, and the number will be contested precisely because it worked.