Theory of Constraints Handbook - James Cox Iii [176]
The current practice in production management inter-mixes MTO and MTS. Economic Order Quantities (EOQs) lead production planning to fill the demand for current customer orders and then add stock intended to cover future orders. This combination of customer orders and stock orders executed in a material requirements planning (MRP)2 environment uses the problematic notion of the “available to promise” algorithm. This algorithm helps in deciding whether current requests can be reasonably met in quantity and time. The problems with this algorithm3 are twofold: the first is the unreliable way uncertainty in the shop floor is handled, and the second is inconsistency due to the varying levels of stock that is not already assigned to firm orders. From the potential customer point of view, sometimes an order is delivered very soon and sometimes an order is delivered relatively slowly. This is problematic because there is no standard for the customer to rely on.
What makes the mix between orders and stock even more confusing is that in MRP every pass from one level to the next level in the bill of materials (BOM) has its own work order, which often merges the requirements from several customer orders and then inflates the work order even more by adding items for stock. As the expected customer demand changes at the top level, those fluctuations are then exploded to the lower levels in the BOM structure with each new iteration of MRP (often done weekly) thus impacting the ratio between the parts that are required for firm orders and parts that are for stock. This means how much component stock for future parts has been added is arbitrary and not derived based on a calculated decision to maintain a certain level of stock of a specific component. In this way calculating the “available-to-promise,” looking for the available stock of a large number of components, is very tricky indeed. It could easily be that for a certain end product some of the required components have a lot of stock, while other components are short. MRP developers have tried to treat the effect of this nervousness by providing pegging (Blackstone, 2008, 97 ) “to determine requirements traceability, which allows one to trace the source of requirements through record linkages.” (© APICS 2008, used by permission, all rights reserved.)
Another source of confusion is the reliance on forecasting, or rather the common misunderstanding of how to use forecasts to support good decisions.
The Common Misunderstanding of Forecasts
The forecasting algorithm is not a prophecy and was never intended to answer questions like, “How many units will be sold next month?” Forecasting is a statistical model that describes, under certain assumptions, a specific uncertain future behavior of a specific variable. Being just a statistical model means all it can do is point to a possible spread of results treated in a solid statistical way—finding a probable average and a probable standard deviation around that average. By providing this partial information on the possible range of results, it allows the decision maker to consider where in the range it is best to place the quantity in question for minimum risk.
The common misunderstanding of forecasting has two parts. The first is to understand what partial information the forecast should provide. The second is how to make a good decision based on the forecast information.
The common ignorance regarding the first part is focused on using the forecast as a single number. The mathematical/statistical handling of all uncertain functions includes, at the very least, two parameters. The common minimum description of uncertain behavior is the use of the average and the standard deviation. Another option is to describe a spread of possible results by the confidence interval: a range of results that encompasses, according to the forecasting assessment, 95 percent or more of the possible results.
The common use of the forecast as a single number is causing huge confusion because the essential range of results is missing.