The page's own worked example — 10 parts, 3 operations, 1 minute per part per operation. Batch of 10: op 1 runs all ten (minutes 0–10), op 2 all ten (10–20), op 3 all ten (20–30); every part gets 3 minutes of work and 27 minutes of queue, and all ten leave together at minute 30. One-piece flow: part n starts at minute n−1 and leaves at minute n+2, so the first is out at 3 and the tenth at 12. First part 10× sooner (30 → 3). All ten 60% sooner (30 → 12). Part-minutes spent inside the cell: 10 × 30 = 300 batched against 10 × 3 = 30 flowing — the 90% WIP cut, countable rather than calculated, because any vertical line crosses ten bars in the top panel and three in the bottom. One caveat on the table's '3–12 min' flow lead time: that is measured from the release of the order of ten. Measured from when the cell starts on a part — what the lower panel draws — every part takes exactly 3 minutes; the 3-to-12 spread is the staircase.
The Batch Illusion
Batching feels efficient. A CNC operator sets up once and runs 50 parts. A painter loads 30 assemblies into the booth at once. A kitter pulls material for 20 units and stacks them on a cart. Each individual operation looks productive — the machine is running, the painter is painting, the kitter is kitting.
But what happens between operations? Those 50 parts sit in a queue waiting for the next machine. The 30 assemblies wait 2 days for the paint to cure because they all need to be done before the booth opens for the next batch. The 20 kits sit on the cart for a week because the line can only consume 4 per day. Batch efficiency at the operation level creates massive waste at the system level.
Scenario: 10 parts, 3 operations, each 1 minute per part.
| Metric | Batch of 10 | One-Piece Flow |
|---|---|---|
| First unit done | 30 min (must wait for all 10 at each station) | 3 min (flows through all 3 stations) |
| All 10 done | 30 min | 12 min (3 + 9 more × 1 min each) |
| WIP at any moment | 10–30 parts | 3 parts (one at each station) |
| Lead time per part | 30 min | 3–12 min |
| Time to detect defect | After 10+ parts processed | Immediately (next station catches it) |
Same work content. Same machines. Same operators. But one-piece flow delivers the first part 10× faster, reduces WIP by 90%, and detects defects immediately instead of after 10+ units are affected.
Cell Design Principles
One-piece flow requires physically co-locating sequential operations into a cell. The five cell design principles:
Arrange Operations in Process Sequence
Not by machine type (all CNC together, all deburr together) but by product flow (CNC → deburr → inspect for this part family). This eliminates transport between operations.
Use a U-Shape Layout
The U-shape puts the first and last operations next to each other, allowing one operator to manage both ends. It also minimizes walking distance and creates visual line-of-sight across all stations. An operator in a U-cell can see the entire process at once.
Balance Work Content to Takt
Each operator’s work content should be close to (but not exceeding) Takt time. Use a Yamazumi chart to visualize balance. An operator can run multiple machines if their combined cycle time fits within Takt.
Design for Flexibility
Cells should accommodate 1–N operators depending on demand. At low demand, one operator walks the entire cell. At high demand, each station has a dedicated operator. This is why U-cells work: the number of operators flexes without changing the layout.
Eliminate Monuments
A “monument” is a large, immovable machine that serves multiple product families and prevents cell formation. The lean response: right-size equipment (smaller, dedicated machines), reduce changeover time so shared equipment can serve the cell economically, or use a shared machine with FIFO lanes feeding it.
Operator Balance Chart
The operator balance chart shows how work is distributed among operators in a cell, compared to Takt time. It is the companion to the Yamazumi chart but focused on people rather than stations.
Cell: 5 operations, Takt time = 60 seconds. Total work content = 150 seconds.
Minimum operators: 150 ÷ 60 = 2.5 → round up to 3 operators.
| Operator | Operations | Work Content | % of Takt |
|---|---|---|---|
| Operator A | Op 1 (25s) + Op 2 (30s) | 55 sec | 92% |
| Operator B | Op 3 (28s) + Op 4 (22s) | 50 sec | 83% |
| Operator C | Op 5 (45s) | 45 sec | 75% |
Balance efficiency: 150 ÷ (3 × 60) = 83%. Operator C has slack — this is where you look for improvement opportunities to potentially eliminate one operator position (reduce total work content to ≤120s and run with 2 operators).
When One-Piece Flow Is Not Possible
Pure one-piece flow is not always achievable. Common barriers in aerospace:
| Barrier | Example | Lean Response |
|---|---|---|
| Cure/process times | Sealant cures 4 hours, paint cures 8 hours | Design “continuous flow with planned buffers” — flow between cure steps, buffer during cure |
| Shared equipment | One autoclave serves 6 product families | FIFO lanes feed the shared resource; SMED to minimize changeover; schedule in product family sequences |
| Long cycle times | Aircraft position takes 5 working days | Flow at the station level: work packages flow within the position on a Takt pitch. See Takt Time. |
| Very different cycle times | Op 1 = 2 min, Op 2 = 45 min | One operator runs Op 1 for multiple cells. Op 2 runs at its own pace with a FIFO buffer between them. |
The principle remains: move toward smaller batches and fewer queues, even when one-piece flow is not achievable. Going from batch-of-50 to batch-of-5 captures 80% of the lead time improvement. Going from batch-of-5 to one-piece captures the remaining 20% — pursue it, but do not let “we can’t do one-piece” prevent you from shrinking batches.
🎯 The Bottom Line
One-piece flow is the ideal: each unit moves directly from operation to operation without waiting. The math is unambiguous — flow beats batching on lead time, WIP, quality feedback speed, and flexibility. Implement it through U-cells designed around process sequence, balanced to Takt time, with flexible staffing for varying demand. Where pure flow is not possible, pursue the smallest practical batch size and use FIFO lanes to connect non-flow operations. Next: Pull System Architecture — designing the signal system that controls when and how much to produce.
Interactive Demo
Run the same parts through the same steps, batched and one at a time. The processing work is identical; only the lead time changes.
Stop reading, start modeling
Model your process flow, run simulations, optimize staffing with TOC math, and test your knowledge with 107 interactive checks — all in one platform.