# Capacity, utilization, and the economics of throughput
A plant manager stares at a spreadsheet at 6 a.m. Orders are backing up. Two options sit on the table: run a second shift (higher labor cost, no capital outlay) or buy a new mold that cuts cycle time (large upfront spend, permanent gain). The finance team wants a number. The floor supervisor wants a decision. Both answers hinge on one question most people get wrong: how much can this plant actually make?
That is the difference between theoretical capacity and real throughput. Let us model it.
Capacity is the maximum output a resource could produce if everything went perfectly. A press that molds one part every 30 seconds has a theoretical capacity of 120 parts per hour.
Utilization is how much of available time you actually use the resource. If the press runs 16 of 24 hours, utilization is 67 percent.
Throughput is the rate of good units the whole system delivers to the customer. This is the number that pays the bills. Not parts started. Not parts molded. Good parts shipped.
The trap: managers optimize utilization (keep every machine busy) when they should optimize throughput (keep the system flowing). A machine running full tilt to build inventory nobody ordered is expensive motion, not value.
Overall Equipment Effectiveness (OEE) is the standard metric for how well a piece of equipment performs against its potential. It multiplies three factors:
OEE = Availability x Performance x Quality
Say a press is scheduled for a 480-minute shift.
OEE = 0.875 x 0.909 x 0.97 = 77.1%
A commonly cited benchmark is that "world-class" OEE sits around 85 percent, though this varies by process and should be treated as a rough reference, not a hard target. The point is not the label. It is that this press wastes roughly 23 percent of its potential, and now we know exactly where: setups, small stops, and scrap.
For a clean primer on the math and the six big loss categories, see this OEE overview from OEE.com.
Here is the rule that saves money: you can only increase throughput by improving the bottleneck. Every other improvement is theater.
The bottleneck is the resource with the least capacity relative to demand. In our plant, imagine the flow:
1. Injection molding (5 presses)
2. Trimming and deburring (2 stations)
3. Assembly (3 cells)
4. Packaging (1 line)
Suppose demand is 10,000 parts per day and each stage can produce:
| Stage | Daily capacity |
|---|---|
| Molding | 12,000 |
| Trimming | 9,500 |
| Assembly | 11,000 |
| Packaging | 14,000 |
Trimming is the constraint at 9,500. The whole plant ships 9,500 per day no matter how fast the presses run. Buying a sixth press does nothing. Running molding a second shift just piles work-in-process (WIP) in front of trimming.
This is the core insight of the Theory of Constraints, popularized in Eliyahu Goldratt's book *The Goal*: an hour lost at the bottleneck is an hour lost for the entire plant. An hour saved anywhere else is a mirage.
Changeovers matter most at the constraint. If trimming loses 45 minutes per changeover and runs 6 changeovers a day, that is 270 minutes of lost throughput on the one resource that limits the whole plant.
Reducing setup time is called SMED (Single-Minute Exchange of Die), a method for cutting changeover to under ten minutes by converting internal steps (done while the machine is stopped) into external steps (done while it still runs). Prepping the next mold, staging tools, and pre-heating dies before you stop the machine can recover much of that lost time at near-zero capital cost.
Cut trimming changeovers from 45 to 15 minutes and you may free 180 minutes a day on the constraint. At a 30-second effective cycle, that is 360 extra good parts per day. Free capacity, no new tooling.
Now the 6 a.m. question. Demand has risen to 11,000 parts per day; the plant ships 9,500. The gap is 1,500 parts short daily.
Both options must attack the trimming constraint. Anything aimed at molding is wasted.
Reversibility is the quiet advantage. If the order surge is seasonal or a single large contract, you do not want a permanent asset you cannot unwind.
Compare on contribution margin per bottleneck hour, not on machine cost.
Contribution margin is price minus variable cost per unit. Multiply by the extra good units each option delivers, then subtract the option's own cost.
A practical rule of thumb: if the demand increase is uncertain or short-lived, favor the reversible second shift. If demand is durable and the tooling permanently lowers cost per part, the capital investment usually wins because its benefit never stops.
And always run SMED first. Recovering 360 parts a day for the cost of a staging cart and a checklist can shrink or erase the gap before you spend on either option.
Knowledge check
1. Why does the lesson argue that maximizing utilization can be a mistake when the real goal is throughput?
2. A press molds 120 parts in an hour, but 10 are scrapped and 30 sit unsold in inventory. Which figure best represents throughput?
3. Time lost to a machine changeover (setup) primarily reduces which OEE factor?
4. Select ALL correct answers about how the three OEE factors relate to different loss types.
Select all the correct answers.
5. Select ALL correct answers that correctly distinguish capacity, utilization, and throughput.
Select all the correct answers.
Two traps catch smart teams.
The bottleneck moves. Add a second shift to trimming and suddenly assembly (11,000 capacity) becomes the new constraint above 11,000 units. Model the next bottleneck before committing, or you will solve one problem and immediately buy another.
Utilization theater inflates WIP. If you reward supervisors for keeping presses busy, they will overproduce upstream of the constraint. That WIP costs cash, floor space, and quality (parts age, warp, or get damaged). Throughput accounting says: measure the plant on parts shipped and on inventory held, not on how busy each island of equipment looks.
A useful discipline is to schedule the whole plant to the pace of the constraint (a "drum-buffer-rope" approach): the bottleneck sets the drum beat, a small buffer of WIP protects it from starving, and a rope holds back material release upstream so inventory does not balloon.
You can sketch the whole decision on one line before opening a full model:
Effective capacity = Theoretical capacity x OEE
Throughput = min(Effective capacity across all stages)
Gap = Demand - Throughput
Best fix = lowest cost-per-recovered-unit action AT the bottleneckRun this before any capital request lands on the finance team's desk. It reframes the conversation from "buy us a machine" to "here is the constraint, here is the gap, here is the cheapest way to close it."