Operations
The efficiency trap on the shop floor

Ask a production manager how he judges a good day on the shop floor, and the answer usually comes down to one word: efficiency. Machines running, operators busy, set-ups kept to a minimum, output per shift as high as possible.
It seems obvious. Expensive machines should be used fully. Idle time is waste. And according to cost accounting, the more units a machine produces, the more its overheads are spread out, so the cheaper each unit becomes.
But if efficiency is so good, why do the most “efficient” plants so often have the worst delivery performance, the highest work in progress and the most chaotic month-ends?
Two kinds of work centres
To see what happens, it helps to look at a typical custom-manufacturing plant in two parts:
- Component work centres (machining, fabrication, painting) serve many orders at once. They have a choice: they can pick which job to run next.
- Completion work centres (assembly, packing, dispatch, site installation) work on one order at a time. They have no choice: an order can’t be assembled until all its parts are there.
Component work centres usually sit early in the flow, completion work centres at the end.
How the trap works
Step by step, here is what efficiency thinking does:
- Component shops batch across orders. To save set-ups, a machine runs the same part for five different orders together, then the next part, and so on. Locally, this looks efficient.
- Orders get out of step. Each machine batches differently. By the time parts reach assembly, an order has 80% of its parts, but not the last few. Another order’s parts were made three weeks too early.
- Assembly can’t wait. Near month end, or when the customer starts calling, assembly sends urgent requests: “We need these three parts for order 4512, today.”
- Component shops swing to the other extreme. They break their batches, take unplanned set-ups and run single pieces for one order.
- Capacity is lost on all those extra set-ups, which delays other orders, which triggers more urgent requests.
Do you see what has happened? By trying to save set-ups at the start of the month, the plant ends up taking far more set-ups at the end of it. The pursuit of efficiency destroys efficiency.
Put rough numbers on a single machining cell. In the first three weeks of the month, by batching parts across orders, it takes about 15 set-ups a week instead of 25. That saves 30 set-ups, at about an hour and a half each, roughly 45 hours. Then, in the last ten days, assembly’s urgent requests force about 50 unplanned set-ups for single orders: 75 hours. The cell has lost 30 hours of capacity in the month, while its efficiency report for the first three weeks looked excellent. And those 30 hours were lost exactly when the plant needed them most.
People who have lived through this cycle for years naturally conclude: “Whenever we focus on completing orders, output drops.” They are right about what they’ve seen, and wrong about the cause. Output drops because of the rush, and the rush is caused by the batching that came before it.
The middle path
If batching everything is one extreme and rushing single orders is the other, the answer is a stable middle zone. How do we get there?
The reason work centres can batch freely is that they can see, and pick from, a huge pile of orders. Reduce the pile, and the problem shrinks by itself.
This is the first step of Drum-Buffer-Rope, often called choking the release:
- Cut the time between material release and due date to roughly half of today’s lead time. (In most plants, actual work is less than a tenth of the lead time, so this is safe.)
- Don’t release material before its release date.
Now each work centre has fewer orders in front of it. There are still enough to combine similar jobs sensibly, but not enough to run far ahead on some orders while others starve. Orders arrive at assembly complete. Urgent requests fall dramatically, and with them the unplanned set-ups.
Add a simple colour priority for each order (green, yellow, red, based on how much of its time has been used) and every work centre knows what to run next without needing a schedule from planning.
Plants that do this typically find that the total number of set-ups in a month goes down, not up, and output goes up.
So should efficiency measures be abolished?
Not necessarily, and here is a subtle point.
Non-constraints don’t need to be maximised, but they can’t be ignored either. Their spare capacity is what protects the constraint and the due dates. If a non-constraint’s spare capacity is shrinking, we want to know. Local efficiency numbers can tell us that.
There is also something more basic at work. People batch by instinct, measures or no measures. Give anyone forty gift hampers to pack, and most will lay out all forty boxes, then put the sweets in all forty, then the cards, then wrap all forty. Nobody told them to. Grouping similar steps just feels efficient.
That is why removing the measure alone rarely changes behaviour. What changes behaviour is a process that removes the opportunity:
- Choking the release ensures there aren’t forty hampers to batch, only the ten that are needed first.
- Daily buffer management (reviewing red and yellow orders every day) catches slack effort early, instead of at month end. In a plant where everyone can see the reds, nobody wants to be the reason for one.
The antidote to the efficiency trap is not a new measure. It is a process that makes the wrong behaviour impossible and the right behaviour visible.
What to do with the old measures
Don’t start the change by fighting over measures. The argument will consume months, and inertia will win it.
Start instead by putting the two processes in place (release control and buffer management) and let results speak. Remove only those parts of the old measures that directly block the change, such as a cap on output that becomes a self-fulfilling target. Once flow improves, the old efficiency numbers usually improve as well, and they quietly stop driving behaviour. At that point, they can be retired without a fight.