Entry 0197·November 17, 2026·Scheduling·Throughput, Variability & Constraints

Who Is Allowed to Move Next Week's Schedule

There is a question with a short answer in a well-run plant and no answer at all in most of them: after next week's production schedule is published, who is allowed to change it, and where is that change written down?
Truth · observed pattern

Who is allowed to move next week's schedule

There is a question with a short answer in a well-run plant and no answer at all in most of them: after next week's production schedule is published, who is allowed to change it, and where is that change written down?

In one multi-line food plant, the published schedule was re-cut late each afternoon. The manager who owned production planning was candid about the mechanics. Holding the plan was that role's job on paper. In practice, a request with more weight behind it would arrive and the plan would move, and nobody recorded the change because the change was normal. That is not a planning failure. It is an authority failure with a planning symptom, and it decides what the plant is able to know about itself.

A moving schedule erases the evidence the plant needs

The schedule is not only a commitment to customers. It is the plant's measurement instrument. Rate, yield, scrap, crew size, and changeover time are properties of a product on a line, not properties of a plant. To produce a number you can defend, you have to know which product runs on which line at which hour, far enough ahead to put an observer, a scale, or a stopwatch there. A plan that resolves at five in the afternoon gives you roughly fifteen hours of notice and no second attempt.

That is where short runs do their real damage, and it is not the changeover clock. More products in the book means each one runs in narrower windows, which means the schedule has fewer places to absorb a late change, which means the change lands on whichever line was going to be observed. One line in that plant was scheduled to run a particular product on two days of a given week. Miss both and the study moves a week, and the following week depends on a plan nobody has written yet.

The loss compounds, because the measurements that survive a moving schedule are the ones that do not care which product ran. Those are the plant-level averages. So the plant accumulates aggregate scrap and aggregate downtime and loses every product-specific standard. This is why a plant can hold a complete demonstrated-rate study from its equipment makers, covering every line and every product, and still be unable to say what a shift ought to produce. The machine rate was established under controlled conditions. The schedule it runs under is not controlled at all.

The record of loss degrades with it. When a stop happens inside a day that has already been re-planned, the surviving artifact is a note on a whiteboard with no start time and no duration, because the shift's attention went to recovering the plan rather than describing why it broke.

Give the schedule a holder, a window, and an exception log

The decision here is not which scheduling tool to buy. It is an authority decision with three parts, and all three have to land together.

First, define the freeze window: the forward period inside which the published schedule is fixed. Its length follows your shortest meaningful changeover and your longest material lead time, not anyone's preference. Second, give the window a holder with standing. One person holds it, and their refusal has to survive a senior request, which means the person who can overrule them is named in advance and the escalation is deliberately slower than the request. A holder who can be overruled in a hallway is not a holder. Third, log every in-window change: who asked, who approved, the reason, and the time. Count the log weekly. That count, not the tool, is the control.

Then reconcile. Compare the schedule you published against the schedule you ran, at the line and product level, for the frozen window. These are three different states and they are routinely treated as one: a schedule you published, a schedule you ran, and a schedule you can reproduce on demand. Only the third supports a standard.

Ownership usually belongs with supply chain or production planning, because that function already carries the forecast and the material commitments. The test is behavioral, not organizational. If sales or a plant leader can change the plan without a logged approval, that person owns the schedule, whatever the chart says, and the planning function is doing clerical work.

This is also a sequencing rule for anyone quantifying value. I do not quote a labor case or a line-rate case off a plant whose schedule is re-cut on the day, because the study that produces the number cannot run under those conditions, and a number that cannot be reproduced will not survive the first challenge from a CFO. Scope the freeze first. It is cheaper than the study it protects, and it is a precondition for that study rather than a parallel workstream.

What a held schedule looks like from the outside

Published and actual schedules reconcile to line and product for the entire frozen window. In-window changes are counted every week, each with a named approver and a written reason, and the count is reported to the same people who see service level. Every downtime event carries a start time, a stop time, and a reason code, so a stop can be priced rather than remembered. A time study booked two weeks ahead runs on the day it was booked. Crew standards are stated per product and per line, not per plant.

The test to run this week

Pull last week's published schedule and last week's actual runs. Count the lines whose product assignment changed after publication. For each change, find the approver and the reason in writing. The changes you cannot attribute are the real number: that is how many times the plant overruled its own plan with no record of who decided or why. Fix the attribution before you commission a study, sign a capital case, or set a crew standard off that plant's data.

Published November 17, 2026
Related reading in Throughput, Variability & Constraints