How to build a production schedule

Outcome

You have a production schedule that says what runs where and when. The shop floor can read it without interpretation, and it survives contact with the first change.

What a schedule has to contain

A production schedule answers one question, again and again: this job — where, and when? Every entry therefore needs four things.

The job. What you make, in enough detail that the floor recognizes it.
The place. The workstation it runs at: the mill, the bay, the bench.
The window. The start and the end. For most shops, dates are the right granularity.
The claim. How much of that workstation it consumes while it is there.

Priority, customer, due date, and operator are all useful. Those four items are what makes it a schedule and not a list.

Step 1 — List the places before the work

Write down every workstation that the work passes through. For each one, record what stands at it and how much it can take in a day.

Do this before you touch the job list. A schedule assigns work to places. If the places are vague, nobody can check the assignment.

This step also decides whether your schedule can catch its own mistakes. You cannot detect that two jobs collide without a shared definition of what they collide over.

Keep machines and their locations as one thing, not as two lists. The mill is a workstation, and what stands at it is a property of that workstation. Two separate lists drift apart, and the schedule then trusts the one that is wrong.

Step 2 — Get durations you believe

For each job, record how long it takes at each workstation, and how much of the workstation it takes per day.

Use what your shop can estimate honestly. Historical actuals beat quoted times. Quoted times beat guesses. A rough number you believe beats a precise number you do not believe.

Note the setup separately when it is significant. A two-hour setup on a four-hour job changes which sequence makes sense.

Step 3 — Place the constrained work first

Start with the workstations that constrain output. Start also with the jobs whose dates a customer commitment fixes. Everything else has more freedom than it appears to have, and freedom is easier to spend later.

The common mistake is to place the loose work first. The flexible jobs fill the bottleneck. The committed job that arrives next then has nowhere to go, and you must unpick the plan.

Step 4 — Check the plan against capacity

For every workstation and every day, compare what you placed against what it can absorb. Wherever the booking exceeds the capacity, you did not finish the schedule. You wrote down a wish.

Resolve each overbooked day deliberately. Move the job, add capacity, move it to another workstation or site, or renegotiate the date. There are four options, and a choice of none of them is itself a choice, made silently.

Capacity planning covers this comparison in more depth. The free template does it by formula and colors the overbooked days red.

Step 5 — Publish it where one version exists

The people who do the work must be able to read the schedule. There must be exactly one schedule.

This sounds trivial, and it is the most common point of failure. A schedule that lives as a file on one laptop, printed on Monday and annotated by hand through the week, is three schedules by Thursday. Whatever you use, apply this test: if two people look right now, do they see the same thing?

Step 6 — Plan for the change, because it is coming

The real quality of a schedule is not how good it looks on Monday. It is how cheaply it absorbs the rush job on Tuesday.

When a job moves, everything it consumed moves with it. Anything that no longer fits must resurface as a decision.

If that reshuffle is manual, keep the schedule coarse enough that a rebuild takes minutes and not an afternoon. Plan at the resolution your estimates support. A 20-minute saw operation needs minutes. A three-day mill job does not.

Set a rhythm for the review: a fixed weekly review, plus a re-check after anything that invalidates the plan.

Three ways schedules stop being believed

Precision that nobody trusts. A schedule to the minute, built on estimates good to the half-day, is fiction with decimal places. The floor ignores it the first time reality diverges, and then ignores it permanently.

Planned to exactly 100%. A shop with no slack cannot absorb the normal. Every rush job then breaks the plan instead of fitting into it.

Updated late. A schedule that lags reality is worse than no schedule. People act on it before they discover that it is stale. If it is too expensive to keep current, the mechanics are wrong. That is a specific, recognizable point, not a failure of discipline.