Cleaning Takes Hours the Capacity Plan Never Counted

Every Monday the same question lands on the planner's desk: how much can the line really make this week? The answer comes from a capacity spreadsheet built on line speeds. Running hours times rate, and the number falls out. By Friday the week has come up short again. The cleanout cycles between product classes were never in those numbers. When a line switches from one product class to another, it stops producing while residues are rinsed and surfaces are cleaned. That stop is a real duration: the full cleaning cycle, pre-rinse through final rinse, entered as a single figure. Its length is plant-specific, and it belongs to the production sequence, because which product class follows which decides what has to be cleaned. Cleaning is machine time, and a capacity plan that leaves it out answers the Monday question wrong.

The capacity spreadsheet has no row for cleaning

The run-rate plan multiplies line speed by run time and calls the result capacity. Cleaning enters that plan, if at all, as a fixed allowance or a rule of thumb, the same number whatever the products and however the sequence runs. In the scheduling literature that omission is the norm rather than the exception: most published scheduling models treat setup time as negligible or fold it into processing time, a simplification known to hurt exactly the plans that need explicit setups. None of this indicts any one plant; it describes what the models assume. The consequence is a capacity statement built from run rates, with no row for the cleanout between product classes. Its answer to the Monday question overstates capacity by exactly the time that was never entered. Those cleanout hours recur, and they consume machine time the plan never shows.

Cleaning is capacity, not overhead

Overhead is a lump you allocate and forget. Capacity is time a machine actually has. A cleanout between product classes occupies a machine for a real, sequence-dependent duration. Allergen, flavour, colour and potency boundaries decide which clean sits between which products, and switching to a different flavour family demands a deeper clean than running the same class again. Every hour that clean runs is an hour the machine does not produce. Call cleaning overhead and it lives outside the plan, an allowance to be absorbed. Call it capacity and it has to be scheduled, like every other hour of machine time. Cleaning between product classes is a scheduling decision: it has a place in the production sequence and a duration the schedule has to show. The rest of this post covers the two ways it can be scheduled and what changes once it is.

Two mechanisms, one planner's heuristic

The vocabulary comes from two documented mechanisms, and both are planner-entered inputs.

The first is a regular cleanout between product classes, entered as a changeover: a cleanout time between the two product classes, on the machine that must clean. Because it belongs to the product-to-product transition, it moves with the production sequence. Change the sequence and the cleanout load changes with it.

The second is a longer clean-in-place (CIP) cycle, a deep clean, or a clean that falls due on a run-hours or calendar interval, booked as a planned downtime window: machine unavailability the planner places on the calendar.

The heuristic that maps them is the planner's. If the cleaning is a recurring between-class cleanout driven by the sequence, enter it as a changeover. If it is a longer or one-off clean, book it as a downtime window. That choice is a modeling decision between two documented mechanisms, never an algorithm decision about when cleaning happens. The schedule models what the planner enters; it does not decide that a clean is due.

Every claim here stays per machine, because Schantt has no shared-resource entity distinct from the machine: each machine's availability is scheduled on its own. In many plants a single cleaning skid serves several machines and can clean only one of them at a time. For machines sharing a skid, the planner staggers the downtime windows by hand and confirms the result on the Gantt. The coordination is the planner's, not the software's.

And the durations stay ranges, plant-specific. The same cleaning cycle runs quickly against a lightly soiled surface and far longer where soil is encrusted, so no fixed per-product allowance can be right for every clean. The planner's own values are the ground truth.

What a schedule that includes cleaning can do

Three things change once cleanout time is in the schedule.

First, the sequence can trade cleanout load. The schedule folds changeover time between a machine's consecutive jobs into its timing, and therefore into total production time. Plans that keep similar product classes together carry less cleanout time, and the schedule prefers them. In Auto mode it can reorder the sequence toward a lower-cleanout arrangement. In Semi-Auto it holds the planner's sequence and lets the machine assignment carry the changeovers instead. The entered durations never change: the schedule trades the sequence and the assignment against them, never the clean time itself. Plants already run batches of one product class together to spread the cleaning across a longer run, and changeover-aware sequencing produces that same clustering on its own, with no grouping rule entered anywhere in the model.

Second, the time is visible. A changeover renders as its own labelled segment ahead of the processing bar. A booked cleaning window appears as a shaded overlay, with detail on hover. The cleanout stops being an invisible allowance and becomes a bar on the Gantt the planner can see, question, and move.

Third, the answer to the Monday question changes. Changeover time lands in each operation's start time and so in the total production time; the schedule subtracts a booked downtime window from working capacity before it plans anything. Between them, a schedule that counts cleaning shows whether this week's job list fits the hours that remain. It corrects the run-rate spreadsheet's optimism down to the machine time that actually exists.

What to change on Monday

For planners, the list is short. Enter regular between-class cleanouts as changeover durations on the machine that cleans. Book longer CIP cycles, deep cleans, and cleans that fall due on an interval as planned downtime windows. On a shared cleaning skid, stagger the windows by hand and confirm them on the Gantt. Then let the schedule trade cleanout load, and read the result.

The schedule is only as grounded as the values entered. Cleanout durations are estimates in practice, and measured plant values should replace them. The model cannot be better than the number it is given.

For buyers comparing scheduling tools, ask whether cleaning is a schedulable decision: entered per machine, visible on the Gantt, tradable against the production sequence. If the tool answers with a fixed allowance hidden in a spreadsheet, it is still treating cleaning as overhead.

Cleaning was always machine time. The only question is whether the plan accounts for it.