Batch vs Flow: the maths nobody shows you

The argument you keep seeing

If you spend any time on manufacturing LinkedIn, you’ll have seen posts arguing that batch production is outdated and one-piece flow is the way forward. The argument usually goes something like this:

On the surface this sounds compelling. And in certain conditions, one-piece flow genuinely is the right approach. But there’s a problem: the people making this argument almost never mention the assumptions their model depends on. When you look at the maths of what actually happens when you connect stations directly together with no buffer between them, the picture changes quite a lot.

The coupling problem

When you run one-piece flow, each station is directly coupled to the stations either side of it. There is no buffer, no decoupling stock, nothing to absorb the impact when something goes wrong at any point in the line.

This means that if any single station stops — whether that’s a breakdown, a quality issue, a material shortage, or just someone needing the toilet — every station upstream is immediately blocked and every station downstream is immediately starved. The whole line stops.

The maths for this is straightforward. If each station runs independently at, say, 90% availability, the overall system availability is not 90%. It’s 90% multiplied by itself for every station in the line:

Four stations at 90% each gives you a system that runs at 65.6%. You’ve lost almost a quarter of your throughput to the coupling effect alone. The more stations you add, the worse it gets.

Try the numbers yourself

Adjust the sliders to see how system availability degrades as you add more stations or reduce individual uptime. These numbers assume one-piece flow with no buffers between stations.

Watch it happen

This simulation runs two identical production lines side by side. Both have 5 stations, and both experience exactly the same random breakdowns at the same times. The only difference is that one has no buffers (one-piece flow) and the other has small buffers of 4 units between each station.

Set each station’s availability and cycle time on the left, then hit start. Both lines experience exactly the same random breakdowns — the only difference is whether there are buffers between stations. Try reducing availability on one station, or giving them different cycle times, and watch what happens. All changes apply immediately — no need to pause or reset.

Three assumptions that don’t hold up

The one-piece flow argument depends on a set of assumptions that are rarely stated, and even more rarely true in real manufacturing environments.

What actually works

The answer is not to go back to massive batches with weeks of WIP sitting between stations. Nobody is arguing for that. The answer is strategic decoupling — small, carefully sized buffers between stations that allow each part of the line to keep running when something goes wrong elsewhere.

This is not a complicated idea. You size your buffers based on the reliability of the stations either side and the cycle time variation you actually observe. A buffer of 5–15 minutes of work between stations is often enough to absorb the vast majority of short stops and minor breakdowns without requiring large inventories.

The result is a system that is dramatically more robust than one-piece flow while keeping WIP low, lead times short, and quality feedback fast. You get most of the benefits the flow advocates promise, without the catastrophic coupling penalty.

When one-piece flow does make sense

None of this means one-piece flow is always wrong. There are conditions where it genuinely is the right approach:

The point is not that one-piece flow is bad. The point is that it is a tool with specific conditions of use, not a universal improvement. Presenting it as always better than batching is not science — it’s dogma.