Lean tools

Flow (one-piece, push, pull)

In French: Flux (pièce à pièce, poussé, tiré)

Flow describes the way work moves from one step to the next. Lean seeks to make it continuous and to trigger it by real demand rather than by forecasts.

Push versus pull

In push flow, each station produces what it can and passes it to the next. The pace comes from a schedule set in advance, often from forecasts. Work in progress piles up in front of the slow stations, and nobody sees it as long as the stock absorbs.

In pull flow, a station produces only when the downstream station has consumed. The signal travels back from real demand — that is the role of kanban. Work in progress is capped by construction, and a disruption becomes immediately visible instead of being diluted in a stock.

Pull flow is not "better" in absolute terms: it is more demanding. It makes problems blocking, which is its point and its difficulty. Deploying it on an unstable process produces stoppages that the stock used to hide, and leads too quickly to the conclusion that the method does not work.

One-piece flow

Handling items one at a time rather than in batches is the most counter-intuitive lever of Lean. Intuition says a batch is more efficient — fewer changeovers, less handling. Lead time says the opposite: as long as the batch is not finished, none of its items moves on to the next step.

Over three one-minute steps, ten parts handled in batches come out after thirty minutes; handled one by one, the first comes out after three minutes and the last after twelve. Same work, same headcount, lead time halved and first result available ten times sooner.

It is this reasoning that justifies SMED: if the batch exists to amortise a long changeover, shortening the changeover removes the reason for the batch.

The bottleneck governs the flow

The throughput of a chain is that of its slowest step, and nothing else. Improving a station that is not the bottleneck does not change the output — it only piles up more work in progress in front of the bottleneck.

The practical consequence is brutal: before any improvement project on a chain, one must know where the bottleneck is. A balancing chart of station times points to it within a few hours, and avoids months of work on the wrong step.

The common mistakes

Measuring efficiency station by station. A workshop where every machine shows a high utilisation rate can have a catastrophic lead time: the high rates come precisely from the work in progress waiting in front of each station. The local indicator rewards what the flow seeks to remove.

Believing that pull flow removes all stock. It caps it and makes it visible; it does not abolish it. A supermarket of components in front of a station with a long changeover is a stock perfectly compatible with pull flow — it is a sized stock, not an endured one.

On the shop floor — Insurance claims processing

A department handled requests in daily bundles, each clerk taking their pile in the morning. Switching to one-at-a-time handling, file by file, with no other change, brought the average lead time down from nine to four days. The number of files handled per day, for its part, did not move.

The other tools illustrated by this trade: banking, insurance and back office.

Frequently asked questions

What is the difference between pull flow and just-in-time?

Just-in-time is the goal — produce what is needed, when it is needed, in the quantity needed; pull flow is the mechanism that achieves it, by triggering production on actual consumption. One can aim for just-in-time without pull flow, but one then falls back on forecasts, and hence on safety stock.

Is one-piece flow applicable everywhere?

No. Some processes impose batches — heat treatment, sterilisation, baking. The useful question is not "batch or one-piece" but "what is the smallest sustainable batch size", and what would have to change to reduce it further.

How do you spot the bottleneck?

Three signs converge: it is the station in front of which work in progress piles up, the one that never stops for lack of work, and the one whose cycle time is the longest. A balancing chart compares station times with the takt time and shows it at a glance.

Should all work in progress be removed?

No — work in progress in front of a bottleneck protects it from upstream disruptions, and removing it would lose throughput where it costs the most. The principle is to place stock where it serves and remove it everywhere else, not to aim for zero on principle.