Executive Guide: Mastering Kanban Card Sizing & CONWIP Flow Control
1. Classical Toyota Kanban Formulation
In lean manufacturing and pull-based supply chains, the Kanban system acts as an authorization mechanism that prevents overproduction—the deadliest of the Seven Wastes (Muda) identified by Taiichi Ohno. Rather than pushing work orders based on forecast projections, work centers only fabricate or transport parts when a circulating Kanban card is freed by downstream consumption.
The required number of Kanban containers ($K$) is determined by the maximum expected consumption during replenishment lead time, plus a strategic safety buffer:
Where:
- $D$ (Average Daily Demand): Expected consumption rate of the part or subassembly.
- $L$ (Replenishment Lead Time): The end-to-end duration required from the moment a card is detached until a replenished container arrives at the staging buffer. This includes card transmission time, queue wait, machine changeover, cycle execution, and transit.
- $alpha$ (Safety Factor): A policy buffer percentage (typically $0.10 le alpha le 0.30$) reflecting the degree of volatility in supplier delivery, machine breakdown (MTBF/MTTR), or demand fluctuations.
- $C$ (Container Capacity): Standard number of units held in a tote, rack, or carrier.
- $lceil cdot ceil$: The ceiling operator, rounding up to the nearest integer card to prevent stockout under-coverage.
2. Little's Law, Queuing Dynamics, and Cycle Time Compression
One of the most profound operational laws governing manufacturing systems is Little's Law, formulated by John Little in 1961:
In an unconstrained push factory, when unexpected bottlenecks emerge, managers often release more work orders onto the floor to keep downstream workers "busy." However, Little's Law dictates that increasing Work-in-Process ($ ext{WIP}$) directly expands manufacturing lead time proportionally, without increasing throughput if the bottleneck is already fully saturated.
By strictly enforcing a Kanban card limit or a CONWIP (Constant Work-in-Process) authorization ceiling, total shop-floor inventory is capped at $K imes C$. Excess orders are held in a virtual digital backlog rather than clogging the physical shop floor. This drains queues, exposes quality defects immediately, and reduces manufacturing lead time by 50% to 80%.
3. The Water and Rocks Metaphor: Continuous Improvement
In Toyota Production System doctrine, inventory is viewed as the level of water in a river, while operational problems (machine downtime, long changeovers, scrap defects, absent workers, unreliable vendors) are rocks hidden beneath the surface. When inventory ($K$) is high, the ship sails smoothly over the rocks without disruption, but the underlying operational deficiencies remain concealed and unaddressed.
Lean practitioners deliberately remove one Kanban card at a time to lower the water level until a process rock is exposed. The engineering and quality team then uses root-cause analysis (Kaizen, SMED, 5 Whys, Poka-Yoke) to grind away the rock before removing the next card.
4. CONWIP vs. Traditional Kanban
| Dimension | Station-by-Station Kanban | CONWIP (Constant WIP) |
|---|---|---|
| Control Scope | Local WIP limits at every buffer between workstations. | Global WIP cap across the entire production line. |
| Part Mix Flexibility | Rigid; each part number requires its own dedicated cards. | High; cards represent work capacity, adapting to high-mix production. |
| Implementation Complexity | High; hundreds of circulating cards and visual boards. | Low; single authorization gate at line entry. |
| Paced Bottlenecks | Buffers paced station by station. | Naturally starves the bottleneck if upstream flow stumbles. |