Finite capacity scheduling is production planning done inside the limits of the plant you actually have: the machine hours available, the changeover times those machines really need, the tooling, the materials on the floor, and the people on shift. Most plants schedule the other way, backward from due dates, assuming every work center will absorb whatever the dates hand it. That plan is a guess made on Friday about a week that has not happened yet. By Tuesday a machine is down, a supplier has shorted a lot, and a changeover ran ninety minutes long, so the plan on the board describes a week that is not happening. The scheduler I remember most clearly kept two plans. One was the printed schedule taped to the wall outside the production office. The other was a clipboard he carried, marked up in pencil, and it was the one the line actually ran to. Nobody thought this was strange. The printed plan was for the meeting. The clipboard was for the plant. Why the plan stops matching the plant by Tuesday A schedule is a set of promises about the future built on assumptions about the present. Every assumption in it stops being true at some point. The plan assumes the wrapper runs at nameplate speed. It assumes a changeover from product A to product B takes forty minutes, because that is the average someone typed in four years ago, and it does not know that A to B is forty minutes while B to A is two hours of washdown. It assumes the material arrives Monday. It assumes five people on the line, when two are on vacation and one is being trained. Each of those assumptions is reasonable on its own. Stacked together across thirty work centers, they produce a plan that was never physically possible. Then Monday happens. The plan does not know. It was built from a snapshot and it holds that snapshot until somebody rebuilds it by hand, which usually means Friday. So the gap between the plan and the plant opens on Monday morning and widens all week, and the people on the floor close it themselves with a hot list, a phone call, and a pencil. That is the part worth thinking about. The plant is not running without a plan. It is running on a second plan, the one in the scheduler’s head, and that plan is invisible, unrecorded, and leaves the building at 3 p.m. with the person who made it. What infinite capacity scheduling actually costs The costs are rarely on anyone’s report, because each one is booked somewhere else. Work in process piles up in front of the constraint, because the plan kept feeding it at a rate it cannot absorb. Overtime gets approved on Thursday to recover a week that was overloaded on Monday. Changeovers multiply, because when the hot list drives sequence, nobody is grouping runs by tooling any more. Expediting becomes a full-time job for someone with a better title than expediter. And the sales team quietly stops trusting the promise date, so they add a week of their own padding, which becomes the new lead time you compete on. The most expensive cost is the one that does not show up as a number. When the published plan is wrong often enough, people stop reading it. Once that happens, nobody runs the plant from the schedule. It is a document that gets printed and ignored. Compare Friday’s schedule to what actually ran by Tuesday This takes twenty minutes and costs nothing. Do it on one line, this week. Friday, print the schedule for that line for the coming week. Date it and put it somewhere it cannot be edited. Tuesday, after the last shift, pull what actually ran. For every job on the printed plan, check three things: did it run on the work center the plan named, in the sequence the plan gave, and inside the shift the plan said. Beside every miss, write one line for the reason. Not a category. What actually happened. Sort the reasons and count them. The biggest group of reasons tells you which assumption in your plan is broken, and it is usually a different one in every plant. If most of your misses say the changeover ran long, you are using one average changeover time where the real time depends on which product follows which, and fixing that one table will do more for your schedule than any software. If most say the material was not there, you are scheduling inventory you do not have. If most say the machine was down, your availability assumption is nameplate rather than history. If most say there was no operator, you are scheduling machines and forgetting that people are a constraint with their own calendar. Do it for four weeks and you will have something better than a schedule adherence percentage. You will have a ranked list of the specific assumptions your plan gets wrong, in your own plant’s words. What finite capacity scheduling changes on the floor None of what follows needs a software purchase first. Name the constraint out loud. In most plants I have walked, everyone can name the bottleneck in one sentence, and yet the plan loads all thirty work centers as though each one mattered equally. Schedule the constraint to the minute and let the rest of the plant serve it. Make changeover a matrix, not a number. From-product to to-product, with the washdowns and the tooling swaps in it. This is often the single highest-value table in a plant and it is usually four years old. Schedule the people and the tooling, not just the machines. A machine with no trained operator is not capacity, it is furniture. Then make the plan recalculate when reality changes, and make the new plan reach the person who has to act on it. A schedule that updates in a system nobody opens is the printed plan on the wall all over again. The recalculation is only worth something if it ends in a decision on the floor. That last point is where most scheduling projects stall. Most plants buy visibility into the gap and stop short of closing it. Seeing that you are three hours behind at the constraint changes nothing by itself. Being told at 9 a.m. which two orders to resequence changes the day. The Tuesday test is yours either way. The gap between the plan on the wall and the plan on the clipboard is already costing you overtime, changeovers, and promise dates. Right now the only complete record of it is in the head of one person who has been doing this a long time, and who will retire eventually. This week: Freeze Friday's schedule for one line, compare it to what actually ran by Tuesday, and write one reason beside every miss.