QM BenchQuality engineering tools

One-piece flow vs batch: a lean simulation you can play with

Two identical lines make the same order with exactly the same processing times. The only difference is how many parts move together to the next station. Press Play and watch the parts, then compare lead time, work in process and how many bad parts are made before a defect is found. Change the stations, add variation, a bottleneck or changeovers, or follow the seven guided lessons.

Line Atransfer batch 20 · WIP now 1 · shipped 0/20
Cut10 sDrill10 sDeburr10 sAssemble10 sShipped
Line Bone-piece flow · WIP now 1 · shipped 0/20
Cut10 sDrill10 sDeburr10 sAssemble10 sShipped

Squares are parts, coloured by batch. Left of a station: waiting in the queue. Inside: in work. Below: finished, waiting for the rest of the batch before moving on.

Result for the order of 20 partsLine A (batch 20)Line B (batch 1)
First part out13 min 20 s40 s
Order finished (lead time of the order)13 min 20 s3 min 50 s
Average lead time per part11 min 45 s40 s
Average WIP (parts in the line)17.63.5
Maximum WIP204
Throughput90 parts / h313 parts / h
Utilisation of the bottleneck (Cut)25 %87 %

Both lines use exactly the same processing time for every part at every station. Total work content: 13 min 20 s.

Gantt chart, line A (each bar = one part at one station)
CutDrillDeburrAssemble013 min 20 s
Gantt chart, line B
CutDrillDeburrAssemble03 min 50 s
Work in process over time— Line A— Line B
01020013 min 20 s
Formulas with the current numbers
  • Little’s law, line A: WIP = TH × LT = 90.0/h × 0.1958 h = 17.63 (simulated 17.63)
  • Little’s law, line B: WIP = TH × LT = 313.0/h × 0.0111 h = 3.48 (simulated 3.48)
  • Equal stations, no variation: batch Q → order lead time = S·Q·c = 4·20·10 = 800 s; one-piece flow → (Q + S − 1)·c = (20 + 4 − 1)·10 = 230 s
  • Bottleneck: Cut with 10 s → maximum output 360 parts / h

Seven guided lessons

  1. Same work, different lead time. Compare a batch of 20 with one-piece flow on the same line.
  2. The first good part. See how fast the customer (or the next process) gets the first part.
  3. Defects found late. See how batch size multiplies the cost of a defect.
  4. The bottleneck sets the pace. Find the constraint and see where the inventory piles up.
  5. Variation creates waiting. See why balanced lines still have queues when times vary.
  6. Little’s law. Connect WIP, throughput and lead time.
  7. Changeovers and the right batch size. See why SMED comes before one-piece flow.

How the simulation works

An order of Q parts runs through a serial line. Each station works on one part at a time. A part may start at a station when the station is free and when its whole transfer batch has been finished at the previous station. Processing times can vary randomly around the mean (coefficient of variation), and both lines always get the same random times, so any difference comes from the batch size alone. Optional changeovers are added before each new batch at every station.

The formulas behind the result

QuantityBatch of QOne-piece flow
First part outS·Q·cS·c
Order lead timeS·Q·c(Q + S − 1)·c
Example S = 4, c = 10 s, Q = 20800 s230 s

With unequal stations the slowest station, the bottleneck, sets the pace: output = 1 / cycle time of the bottleneck. And for every line, Little’s law holds: WIP = throughput × lead time.

What it means on the shop floor

  • Lead time is mostly waiting. In the 20-part batch example a part is worked on for 40 s but spends on average about 700 s in the line. Value-added time is a small share of lead time.
  • Quality feedback is fast in flow. The time from making a defect to finding it equals the lead time between the two operations. Small batches limit the damage.
  • Inventory hides problems. WIP piles up in front of the bottleneck and in batch trays. Lower WIP makes problems visible and shortens lead time (Little’s law).
  • Improve the bottleneck first. A faster non-bottleneck station only builds more WIP.
  • SMED before flow. Long changeovers make very small batches lose capacity. Shorten changeovers, then reduce batch size.

Frequently asked questions

What is one-piece flow?

One-piece flow (single-piece flow, continuous flow) means that each part moves to the next operation as soon as it is finished, instead of waiting for a whole batch. Stations work on consecutive parts at the same time, so lead time and work in process fall sharply.

How much faster is one-piece flow than batch production?

For S equal stations with cycle time c and an order of Q parts, a full batch needs S·Q·c, one-piece flow (Q + S − 1)·c. With 4 stations, 10 s and 20 parts: 800 s against 230 s. The work content is the same; only the waiting changes.

Is this the lean penny game?

It is the same idea as the penny game (coin game, flip-the-coins exercise) used in lean training: the same work is done in batches and in flow, and the time is compared. The simulator adds random cycle times, a bottleneck, changeovers, a defect that has to be found and the formulas behind the result.

What is Little’s law?

Average work in process = throughput × average lead time (WIP = TH × LT). At the same output, lower WIP means proportionally shorter lead time. The simulator shows both sides of the equation for each line.

Why does a smaller batch improve quality?

A defect is usually found at a later operation or at the final test. With one-piece flow the first bad part reaches the test after a few cycles and the line can be stopped. With a big batch the whole batch is made before anybody sees a part: scrap, rework and sorting multiply.

When is one-piece flow not possible?

Long changeovers, very different cycle times, shared machines, ovens or washers that process lots, or long transport distances. The lean answer is to reduce changeover time (SMED), balance the line, move machines into a cell and use small transfer batches where flow of one is not yet possible.

What is the difference between a process batch and a transfer batch?

The process batch is how many parts are made between changeovers. The transfer batch is how many parts are moved together to the next station. You can keep a larger process batch for a machine with a long changeover and still transfer in small batches, which already cuts lead time.

Related tools