The 30% Problem

Walk into any regulated process plant and ask a planner where the day actually goes. They'll talk about batches, campaigns, cycle times, throughput numbers. They'll show you a schedule that lays out reactor occupancy, line allocation, packaging windows.

Then ask what happens in the rest of the day — the time the schedule doesn't lay out.

The pause is the answer.

In a typical pharma, food, or specialty-chemicals plant, somewhere between a quarter and a third of staffed time is consumed by activities that don't appear on the production schedule as first-class work. Cleaning. Changeover. In-process control. Validation. QC hold. These activities are not waste. They are not optional. In regulated process manufacturing they are the cost of being allowed to operate. And yet most production schedulers — the tools, the workflows, the people running them — treat them as invisible.

That gap between what the schedule plans and what the day actually contains is the 30% problem. It is the single largest unaddressed source of friction in process scheduling, and the cleanest place a plant can recover capacity without buying anything.

Production scheduling tools, broadly, were built around one assumption: that the unit of work is the batch, the run, or the order. Everything else is overhead — fixed buffers, coefficients, planning factors. A clean-in-place cycle becomes "+90 minutes." A campaign-to-campaign changeover becomes "+4 hours." A QC hold becomes "we'll release on Tuesday."

That model was reasonable when plants ran a few products on dedicated lines. It breaks down the moment you're running thirty SKUs on shared equipment, where the changeover sequence determines OEE more than the batch sequence does, and where a single missed cleaning verification can hold an entire downstream campaign for a day.

The schedule treats indirect time as a flat tax. Reality treats it as the load-bearing structure of the day.

Four categories show up in almost every process plant. Each one is structurally different. Each one is scheduled badly for the same reason — the production scheduler doesn't have first-class language for it.

• Cleaning and sanitation. CIP, SIP, allergen wash-out, dedicated wash bays. The duration depends on what you ran last, what you're running next, and which validated cleaning cycle applies to the pair. In a multi-product plant the cleaning matrix is its own combinatorial problem. Most schedulers code it as a static pre-batch buffer and hope.
• Changeover. Tooling, recipe loading, line sanitisation, pre-batch inspections. In food and beverage plants the allergen sequence dominates: peanut → dairy → gluten-free is a three-hour set; gluten-free → peanut is a thirty-minute set. The schedule that doesn't see this difference is leaving capacity on the floor every shift.
• Validation, IPC, in-process testing. Cleaning verification swabs. First-article approval. In-process lab samples that gate the next phase. These create blocking dependencies — the line cannot proceed until QC reads "OK" — but they live in the lab schedule, not the production schedule. The two schedules don't talk.
• QC hold. The batch is done; the lab still has to release it. Stability pulls, identity tests, micro hold, dispositioning paperwork. The unit is finished but unavailable. Downstream packaging plans on the unit being available; QC plans on whatever today's analyst load looks like. The two plans collide on Tuesday morning.

A scheduler that names all four of these as events — and re-plans against them as they move — is a different kind of scheduler. A scheduler that calls them "+N minutes" is a forecast.

The cost is in three places, none of which line up neatly with a single KPI.

The first is OEE. Industry benchmarking puts typical OEE in regulated process manufacturing in the 35–65% range, against a world-class TPM benchmark of 85%. The structural gap between the typical and the world-class is dominated by availability losses — and the largest tranche of availability loss in a multi-product process plant is unscheduled or badly-scheduled indirect time. A campaign that finishes 90 minutes late because a cleaning verification slipped doesn't show up as a cleaning problem. It shows up as the next campaign starting late.

The second is planner load. When the schedule treats indirect time as a fixed buffer and reality keeps disagreeing with the buffer, the planner becomes the rebalancing layer. Friday rebuilds. Monday re-rebuilds. Phone calls between production, the lab, and maintenance. Studies of planner-effort burden in process plants are scarce, but talk to anyone running a Monday operations review and they'll tell you a non-trivial share of their planning time is spent reconciling indirect-time assumptions with what actually happened. None of that work moves a batch.

The third is the silo cost. Production, QC, and maintenance all schedule indirect time — production schedules CIP, QC schedules sample analysis, maintenance schedules PMs. None of them sees the full picture. A PM scheduled into a window where QC was about to release a batch is a small, frequent, expensive class of event. It doesn't have a clean line item in any cost report, because by definition it lives between functions.

A scheduler that takes indirect time seriously has three properties.

It models cleaning, changeover, validation, IPC, and QC hold as constrained activities — with their own resources, durations, dependencies, and sequence-dependent rules — rather than as additive buffers on the batch.

It receives events from across the plant in real time. Cleaning verification swab cleared. OOS detected. Analyst out sick. PM brought forward. Each event updates the model; the schedule re-solves against the current state, not yesterday's snapshot of it.

It schedules production, QC, and maintenance against the same plant model — so the cleaning matrix, the lab queue, and the PM calendar are all visible to each other. When QC's release queue extends, production sees it. When maintenance moves a window, QC and production both see it.

The result is not a perfect day. The result is a day where the schedule still describes what's actually happening at 14:00, because indirect time stopped being a fiction in the model and started being scheduled work like any other.

That is what Bodhee Production Scheduling and Bodhee Quality Control Scheduling are built to do — model the indirect activities that schedulers usually ignore, re-solve as events arrive, and schedule across functions instead of through them. Not faster planning. Different planning.

If "we lose a third of the day to things we don't schedule" is recognisable, that is the problem worth solving first.