FIGURE 3-1 The conventional planning schema
MRP was a huge leap forward because for the first time what was required could be calculated based on what was already there compared with what was needed, with the net result time phased. The objective of MRP was to precisely time-phase the requirements and replenishments to dramatically reduce inventory from the previous order point approach where some of everything was kept around all the time. This ability to calculate dependent demand through a bill of material was a significant development. It was no longer necessary to forecast dependent demand—it could be calculated based on the expected demand for the parent part. APICS defines dependent demand as:
Demand that is directly related to or derived from the bill of material structure for other items or end products. Such demands are therefore calculated and need not and should not be forecast. (p. 46)
MRP evolved because of the advent of the computer, and the age of marketing in the 1950s introduced more product variety and complexity than was managed previously. Order point (the previous method of materials management) clearly could not affordably handle these new requirements. To understand how planners deal with MRP on a daily basis, refer to Appendix A, where a simulated environment demonstrates the day-to-day difficulties associated with MRP.
Yet even if a company has 100 percent of the requirements and 100 percent of the assumptions validated, the conventional planning approach will still be ineffective. The remainder of this chapter will explain why.
Distortions to Relevant Information
The conventional planning approach actually creates the bullwhip effect and its inherent distortions to the flow of relevant information and materials. Some of the ways in which conventional planning creates the bullwhip is related to the manner in which convention chooses to use MRP. Other contributions to the bullwhip are related to hard-coded traits in MRP systems. All of these issues, however, are related to one key and fundamental attribute of MRP.
MRP is essentially a calculator. It needs three basic inputs to perform its calculation. One of those inputs is “demand.” Different demand inputs will produce different outputs. The APICS Dictionary defines demand as:
A need for a particular product or component. The demand could come from any number of sources (e.g., a customer order or forecast, an interplant requirement, a branch warehouse request for a service part or the manufacturing of another product. (p. 44)
By this definition, demand can be broken down into two different types: forecasted and actual. Both of the following definitions are from the APICS Dictionary:
The type of demand that is chosen to drive the MRP calculation is a primary determinant of how much relevant information can be produced from MRP. Remember, the flow of information and materials must be relevant to the required output or market expectation of the system. To be relevant, both the information and materials must synchronize the assets of a business with what the market really wants; no more, no less.
A hard-coded trait of MRP is that with a given demand signal, MRP is designed to net perfectly to zero. You make exactly what you need without any excess. In this regard it could be argued that MRP is the perfect JIT system. If the demand signal is perfectly accurate, then the MRP calculation will be perfectly accurate. Given that the math allows no tolerance for error, it seems obvious that MRP should only be given as accurate a signal as possible.
With that in mind, should the demand input to MRP be what a company thinks the market wants to buy or what the customers actually want to buy? Which will produce a more relevant result? As described in the definition of actual demand as well as Figure 3-1, the conventional approach combines both types of demand. Forecast is used to create planned orders, and then demand is adjusted as the picture becomes clearer with actual orders. Why is this problematic?
There are three truths about forecasts:
1. All forecasts start out with some inherent level of inaccuracy. Any prediction about the future carries with it some margin of error. This is especially true in the more complex and volatile New Normal.
2. The more detailed or discrete the forecast is, the less accurate it is. There is definitely a disparity in the accuracy between an aggregate-level forecast (all products or parts), a category-level forecast (a subgroup of products or parts), and a SKU-level forecast (single product or part).
3. The more remote in time or farther out forecasts go, the less accurate they get. Predicting the weather tomorrow is much more accurate than predicting the weather 52 days from today. Yes, history can be used as a basis for a prediction, but the margin of potential error is much higher. It is not uncommon that in many industries the accuracy of a forecast can drop below 10 percent beyond 90 days at the SKU level.
Today many forecasting experts admit that 70 to 75 percent accuracy is the benchmark for the SKU level. Figure 3-2 is the results of a 2012 survey conducted by forecastingblog.com showing the reported forecast error rates across various industries at the SKU level.
Unfortunately, when you start a serial, complex, and interdependent process with an error-prone input, the resulting output integrity must be suspect. Planned orders are derived from these forecasts, and very real commitments of cash, capacity, and materials are directly derived from a prediction that is subject to varying degrees of inaccuracy, sometimes with extremely significant degrees of inaccuracy.
As time progresses, the demand picture changes with the incorporation of actual demand, MRP is rerun, and subsequent changes occur. The result is that we end up with things that we do not need and desperately expedite things we have just discovered that we do need.