The silo tax — a manufacturing-economics argument for cross-functional scheduling

A pharma plant manager once described his Monday morning to us this way: "Production has a schedule. Maintenance has a schedule. The lab has a schedule. Three perfectly good plans, none of which know about each other. The week starts when they meet."

That meeting — the one where three rational plans discover they can't coexist — is what this piece is about. It is the silo tax: capacity that the plant pays for and never recovers, charged not by a vendor but by the operating model itself.

It does not show up on a P&L line. That is precisely why it persists.

This piece is published by Neewee, which builds Bodhee — adaptive scheduling software for production, quality control, and maintenance in process manufacturing. We named this cost "the silo tax" to make it something a CFO can act on.

In any process plant running shared equipment, three planning teams independently optimise against the same physical resources:

• Production scheduling wants to maximise throughput against the order book.
• Quality control scheduling wants to keep release timelines tight: stability pulls, in-process checks, finished-product release.
• Maintenance scheduling wants to honour preventive intervals, regulatory inspections, and equipment-availability commitments.

Each team's KPI is local. Production measures plan adherence. QC measures cycle time. Maintenance measures PM compliance. Every individual KPI can hit green while the plant — the system that all three share — loses capacity to the friction between them.

That friction has a precise economic identity. We call it the silo tax: the difference between the capacity the plant could run if the three plans were jointly feasible, and the capacity it does run because they were optimised separately and reconciled by hand.

The tax does not show up as a single number on a single report. It accrues in five places, all of them operationally familiar and all of them easy to misattribute:

A reactor sits idle because production sequenced around a maintenance window that maintenance later moved. A QC analyst stands at a fume hood waiting for production to finish a lot they thought was already running. A clean room sits empty for ninety minutes between a changeover and a stability pull because nobody held the slot.

These minutes don't end up in the OEE breakdown under "schedule conflict." They get coded as changeover, or wait time, or unplanned downtime, or — most commonly — they don't get coded at all.

Maintenance defers a calibration by 48 hours to honour a production push. Three weeks later, an in-process check fails because the instrument drifted past its window. The lot goes on hold. The QC team reruns the panel. The cost is real. The trace back to "maintenance moved a calibration to help production" never happens because no system links those two events.

When the planner can't predict whether the lab will release a lot on time, or whether the changeover window will hold, or whether the next preventive will land where the schedule says, the rational response is to carry more in-process inventory. That inventory is the planner's hedge against the silo tax. It ties up working capital and absorbs cycle time, but it makes the plant feel calmer. The hedge is so universal that nobody questions it; it just becomes "how we run."

Every regulated plant we have looked at has a recurring meeting — daily, sometimes per-shift — where production, QC, and maintenance manually reconcile their plans. Skilled people, expensive people, spending one to two hours per shift redoing in a room what an integrated scheduler would do in seconds. Multiply across shifts, sites, and a year. That number alone often pays for a scheduling system several times over.

The hardest cost to see is the one that never gets booked. A CDMO turns down a high-margin slot because nobody can confidently say whether the next eight weeks have room. An OSD plant declines a campaign extension because the maintenance + QC + production overlay is too uncertain to commit to. The opportunity-cost line is invisible because the order never enters the system.

The silo tax persists because of a structural property of how plants are organised, not because of any individual failure of judgement.

There is no role in the standard plant org chart whose remit is "the integrity of the multi-function schedule." There are owners of each schedule. There is no owner of the system of schedules. So the cost is real, distributed, and politically homeless.

This is intentionally rough — every plant's number is different, and the point is the order of magnitude, not a defended estimate.

For a mid-sized regulated process plant (100–300 staff, 40–80 SKUs across shared equipment), the silo tax in our experience accrues roughly as:

A plant with €100M of annual capacity that loses 5% to schedule-conflict idle is paying a €5M silo tax in capacity alone, before the inventory and labour line items. That number is rarely visible because the 5% is distributed across a hundred small events, none of which is individually large enough to investigate.

Directional estimates based on Neewee deployment observations in regulated process manufacturing. Individual plant results vary; figures are intended to illustrate order-of-magnitude, not benchmark against.

The honest answer is: not a tool, an operating-model change that a tool then makes feasible. The tool without the operating-model change is shelfware. The operating-model change without the tool is a permanent meeting.

Three things, in order:

The three planning teams need to be working from the same picture of what equipment exists, what state it's in, what changeovers it requires, what cleaning state it currently holds, what operators are qualified for it, and what windows are blocked for maintenance. Today this lives in three or four systems with no shared truth. The first move is to designate the single source of capacity truth and make all three plans consume it. (Bodhee is built around this shared model.)

The point at which production, QC, and maintenance schedules are produced should be the same point. Not three sequential computations followed by a reconciliation pass — one joint computation that respects all three sets of constraints from the start. This is the technical core of cross-functional scheduling. It is also where most "scheduling" tools quietly stop, because joint-feasibility math is harder than sequential-feasibility math.

Adding a system-level KPI changes behaviour faster than any tool deployment. "Cross-function schedule conflicts per week" is a useful one. "Reconciliation-meeting hours per shift" is another. Whichever metric you pick, somebody has to own it and report on it. As long as the only KPIs are local, the local optima will keep colliding.

The silo tax is the largest scheduling cost most regulated process plants don't measure. It is paid in capacity, working capital, labour, and missed orders, and it is invisible because the standard reporting carves the plant into functions that each look healthy on their own.

The cure is not a meeting cadence. It is a single, shared, computed schedule for production, QC, and maintenance — owned by someone whose KPI is the integrity of the joint plan, not any one part of it. The investment case is not "save FTEs." It is "stop paying the tax." That number, when you actually measure it, is large enough to fund the change several times over.