Limiting total work in process with CONWIP

Why it matters

CONWIP is a pull arrangement based on a total work-in-process limit. As work leaves the defined system, authorization opens for the entry of a new job. The shared cards do not have to be tied continuously to a single product. This property can be considered in conditions where products follow different routes and reserving ready stock per product is hard. But the shared limit alone does not determine which product will start first. A separate sequence rule is needed for delivery need, job readiness and resource load. The presence of a card may allow a job to start; it does not by itself prove that all conformity and priority conditions are met. Authorization and selection are kept separate.

A per-product kanban ties the consumption of a defined product to the replenishment of that product. CONWIP controls the total quantity of work by looking at the exit from the shared system boundary. It is not right to carry kanban cards arranged for one product into a variable-route system just by changing their name. The product the card is tied to, the boundary it circulates in and the event that frees it must be redefined. If the same card independently authorizes both the product shelf and the entry to the whole system, it can create a double right to produce. Choosing an architecture is wider than choosing a card form. Unless it is explained where which stock is limited and which consumption starts which production, system behavior cannot be predicted.

The total stock limit does not guarantee that there will be an equal amount of work in front of every station. All the jobs inside the shared limit can pile up at the same resource; other resources may stay idle. The same unit count carries very different workloads in different product mixes. A route change, rework or a long setup can make the cards stay in the system longer. Adding more cards does not reduce this load. In contrast, reducing the limit too much can disturb the feeding of the system when the needed jobs do not arrive in time. The right limit is tested together with capacity and flow conditions. Making the smallest card count or the highest resource utilization the only aim is not enough.

The relation among stock, rate and time is interpreted with steady averages. When the average work in process is divided by the conforming product rate leaving the same boundary, the average time in system can be found. Substituting the card count set as an upper limit for the measured average stock is not always valid. If some cards wait free, the real average stock may be lower. In a temporary period in which the work accumulated at the start is being cleared, entry and exit rates are not in balance. For this reason, before calculating the ratio, the time window, the boundary and the product scope are closed. The calculated average is not a firm promise for the lead time of a specific job; distribution and priority are assessed separately. The figure, example, counterexample and short quiz numbers in this lesson are within the scope of this training. This dataset was generated synthetically for training; it does not represent a real business, employee or customer study.

Formula