Determine the Total Size of the KANBAN Loop

The expression "KANBAN loop" derives from the fact that KANBANs cycle between supplier and customer wprk centers in a type of feedback loop. KANBANs accompany production parts on the forward flow, and cycle back to the supplying work center to advise when the customer center has consumed the stock.

8. The next step involves determining the number of KANBANs required to accommodate the mix of prod­ucts and lot sizes considered in points 1 to 7 above. A certain number of KANBANs is required to justify WIP normally awaiting consumption by the customer work center. Other KANBANs will be accumulated on the planning board until a sufficient number justifies the production of a new lot. The total number of KANBANs in the loop is determined based on the supplying work center's constraints in respect to the consumption patterns of the customer. This analysis was started by targeting the bottleneck work center. A separate loop is determined for each customer-supplier-part rela­tionship.

Since most or all the parts produced by a single work center share similar characteristics, the calculations performed in steps 1 to 7 above can often apply generically to all of them. This is especially true (without being a prerequisite) if lot sizes can remain uniform from one part to another, and where the consumption volumes are generally uniform from one part to another.
Three general aspects must be considered to establish the number of KANBANs in each loop.

8a) First, we have to ensure there are enough KANBANs in the loop to allow the production of a complete lot size of the part being analyzed. For instance, let's assume that KANBANs will be tagged to skids containing 220 units and that the work center produces 250 parts per hour. For a 5.3 hour production lot, we are talking about 1325 parts. Rounding this to 1320 parts will prescribe 6 KANBANs < 1320/220 = 6) for the lot size itself.

8b) Next, we need to implement KANBANs to ensure the availability of stock during the reaction time of the supplying work center to start production on a part whose KANBAN reorder point has been triggered. For instance, in the example above, if 6 KANBANs accumulate on the KANBAN planning board, it would trigger a production lot of 6 KANBANs. However, before this can be executed, in-process lots of another part may need to be completed. So there has to be some inventory (with KANBANs attached) at the customer site to absorb that delay. We also need KANBANs to address the customer consumption lot size if it is significantly larger that of the KANBAN.

Let's start with reaction time. It varies as a function of several factors. Two important ones is the relative size of the lots, and the number of different products in competition for the same resource (work center).

When a part reaches its KANBAN reorder point, the work center will probably be producing some other part. If the average lot size for this work center is 5.3 hours, and setups take 1 hour, the maximum reaction time to changeover will be 6.3 hours (assuming no other part has more priority). The minimum reaction time will be 1 hour, for the setup. The average reaction time relative to the lot size will be about 3.65 hours ([6.5 + l]/2). Stock must exist out there to account for this reaction time. We must then convert these 3.65 hours into KANBANs by estimating how many parts the customer will consume over that period. If, for example, the item has an estimated demand of 3000 per month, this represents 9 per hour (3000/[21 days * 16 hr.] = 9), or 34 over 3.65 hours. If a KANBAN represents 220 items, this represents 0.15 KANBANs. Other issues discussed below will probably justify the rest of the KANBAN; eventually, we will round up to the nearest complete KANBAN (no partial KANBANs are allowed in production).

It is foreseeable that the reaction time may at times be longer than the one calculated above if several parts hit their KANBAN reorder point simultaneously and thus become in competition for the work center's capacity. This should not happen very often, otherwise, it would indicate the work center is overloaded or that the KANBAN loops need to be adjusted to increase the buffers.

The magnitude of the reaction time can be monitored physically if the KANBAN system has not yet been implemented. The in-process queues in front of the work center, measured in work hours, give a good idea of the reaction time. Each "scheduled" order must wait the length of time represented by the queue before being processed. Of course, the implementation of the KANBAN system will naturally shrink the queues 20 to 40% simply because better priority decisions will result. Thus if the queue is 5 days long, use 3 days as an initial competition-based reaction time. In our example, 3 days of consumption represents 432 items, or 1.96 KANBANs (220 each).

To Be Continued

For balance of this article, click on the below link:

Lean Manufacturing Articles and go to Series 01