Making every part of your factory more efficient can make the factory worse.

If you own an expensive machine, you want it running. If you are paying eight people in a department, you want them working. If a machine can produce 100 and it only produces 60, it feels as though you have wasted 40 units of capacity.

That sounds perfectly reasonable.

The problem is that it only makes sense if that resource operates on its own. In a factory, it doesn't.

Go back to the three stations. Cutting can do 100 a day. Printing can do 60. Finishing can do 80. The factory can still only ship 60 finished jobs.

Now suppose we look at Cutting and say: it is only running at 60%. We need to improve its efficiency.

So Cutting produces 100.

What changed? Printing still processes 60. Finishing still receives at most 60. The customer still gets 60.

The factory did not become more productive. It simply created another 40 jobs waiting somewhere between Cutting and Printing.

Those extra 40 jobs are not free. They need somewhere to stand. Someone may need to move them. They need to be counted. They can get damaged. They can get mixed up.

Priorities can change while they wait. An urgent job may now have to fight through a queue of work that did not need to be produced yet.

More work has happened. But the factory has not produced more finished goods.

Keeping a resource busy is not the same thing as making the factory productive. Local activity can go up while finished output stays exactly where it was.

On a real floor this does not arrive as a neat diagram. It arrives as piles of work between departments. Jobs that have been started but not finished. People asking which job is actually urgent. Supervisors walking the line to expedite. Pallet after pallet being shuffled to make space.

Every department can look productive. The efficiency report can look fine. And customer orders are still late — because everybody being busy is not the same as work getting finished.

A road is not successful because it contains the maximum possible number of cars. Once enough cars are added, everybody moves more slowly. The point is movement through the road, not occupancy of every metre of tar.

A factory is the same. The point is finished work leaving the building — not maximum occupancy of every machine.

If Printing can only take 60, then Cutting only needs to supply 60. Cutting will have spare capacity. That does not automatically mean something is wrong.

We have been trained to look at an idle machine or an idle person and see waste. It is an intuitive reaction. A machine that is not running feels like money standing still.

But if making more at that resource cannot increase what the factory ships, keeping it busy for the sake of utilisation may create more waste somewhere else — space, handling, damage, confusion, and a queue that the next urgent job has to climb through.

Sometimes the most useful thing a machine can do, right now, is wait.

That is not an argument for laziness. It is an argument about timing. If producing something now only creates unfinished work that cannot yet become a shipment, waiting is better for the system than producing.

None of this means efficiency does not matter. None of it means utilisation does not matter. And it does not mean people should sit around for the sake of it.

Local efficiency is only useful when it helps finished work move through the whole factory.

If improving a resource means more finished jobs leave the building, that is useful. If it simply causes more unfinished work to pile up somewhere else, we should pause before we call it an improvement.

Why do we behave this way? One reason is that local activity is easy to measure. Machine utilisation. Units per hour. Labour efficiency. Department output. Those numbers are visible. They fit on a report.

None of them necessarily tells us how much more finished product left the factory.

So the question is worth asking plainly. What are we actually trying to improve? A department? A machine? A person's utilisation? Or the flow of finished work through the whole factory?

Those are not the same target. Measuring the first three can make the fourth worse.

We started with a factory capable of Cutting 100, Printing 60 and Finishing 80. We tried to improve Cutting. We made it produce 100. Everybody at Cutting became more productive. The utilisation number improved. The efficiency report probably looked better.

And the factory still shipped 60.

That is the problem with managing a dependent system by measuring its parts independently. The objective is not to make every part move as fast as possible. The objective is to make the work flow through the whole system.

Those are not the same thing.

The usual reply comes quickly, and it is a fair one. We are paying for that machine. We are paying those wages. That machine costs us so much an hour. We need to recover those costs.

Those concerns are real. They also raise the next question: if an hour of capacity at one resource can increase what the factory ships, while an hour somewhere else changes nothing, can we really treat those hours as economically the same?