How a plant actually decides.
A plant’s cost structure is set by four decisions. Each looks self-contained on a spreadsheet. On the floor they interact, and the interactions are where the money leaks. The work is to model the four as one system before the capital is committed, and to read off the configuration the building already supports.
Of everything a plant decides, four move the numbers.
Labor
Labor gets treated as a headcount number on a budget line. On the floor it behaves like a system: which crew, on which line, against which schedule, and the variance between a good shift and a bad one. The plants that win the labor decision staff to the line they actually run, not the one on the nameplate.
The bar: Staffed to the modeled line curve, not the worst piece you will never run.
Automation
Every capital case is a bet about where the constraint will sit after the money is spent. Automate the wrong step and the next constraint downstream eats the savings before they reach the P&L. The automation decision is won before the PO, by modeling the line as it actually runs and reading off where capacity is really gated.
The bar: Every capital case is modeled against the real line curve before the PO, not justified after.
Scheduling
Scheduling is the cheapest capacity in the building, and the one most plants leave on the table. The schedule that looks efficient on Monday strands the floor by Thursday, because changeover and allergen sequencing scale with SKU pairs, not SKU count. Sequencing is a capacity decision wearing a calendar.
The bar: The changeover matrix is modeled and the run sequence is locked to it; the schedule is the capacity plan.
Sourcing
Sourcing is scored on unit cost and lived on the floor. A cheaper film, a new resin, a spec change, a different supplier: each looks like a procurement win and lands as a throughput change the line never modeled. The sourcing decision is a throughput decision wearing a purchase order.
The bar: Every spec and supplier change is scored in floor minutes, not just unit cost.
Every plant decides at one of four levels.
The same four decisions are made at very different levels of maturity. Good is Modeled. Great is Orchestrated.
Reactive
The floor runs on yesterday’s fire. Decisions get made when something breaks, and the plant pays for capacity it cannot see.
Scheduled
There is a plan, but it is built on averages and nameplate rates. It looks efficient on Monday and strands the floor by Thursday, because the average is the one number that hides the stops.
Modeled
GoodThe line is simulated against a year of its own data until it reproduces reality. Every configuration is run virtually before a shift or a dollar moves. The bottleneck is named and quantified, not assumed.
Orchestrated
GreatThe model runs ahead of the line. Labor, automation, scheduling, and sourcing are tuned as one system, continuously, against the demand signal. The factory thinks before it runs.
Every entry names the bar: what the well-run version of that decision looks like, in numbers you can score your own floor against.
A diagnosis tells you a mole exists. A bar gives you a gap to measure. The gap is where the work is, and the gap is what a validated model makes visible.
Adding a crew or a line is the most expensive way to answer a question a validated model answers for free.