Theory of Constraints Handbook - James Cox Iii [198]
The Fallacy of the Variance
A related fallacy involves the understanding of variance. Most forecasting algorithms present the data as an average demand and if one really insists, then the standard deviation is given. The number of people who really understand the meaning of a standard deviation is very limited because this is a mathematical object that does not have any intuitive translation to real-life scenarios. Try to ask a salesperson not just how much he is going to sell, but also what the standard deviation is. This again calls for very sophisticated people interpreting the forecasting results in order to get some benefits from the forecast.
Suppose the salesperson estimates the average consumption of a product at a specific retail location as 17.5 with a standard deviation of 2. How much inventory should be kept at this site? If you stock exactly 17.5 units (assuming that is possible), then you would think you had a 50 percent customer service level. Recall the problem with means stated previously. However, suppose you wanted to satisfy 95 percent6 of the customers requesting this product. How much should you stock? The answer is provided by the following calculations: 17.5 + 1.645(2) = 20.8 units. Stocking only 2 units above the mean (19.5) would provide a customer service level of approximately 86 percent. The critical point is, few people conceptually can estimate a standard deviation and determine its impact on sales without a computer.
The Fallacy of Sudden Changes
Many forecasting methods7 can track changes in demand, but the more sudden the change the worst the forecast will be. An example follows. A very enthusiastic article in a paper has just appeared that suddenly changes the consumption pattern in the whole region. Suppose that the article summarizes a breakthrough study in cancer prevention and stated that if a person drinks one glass of cranberry juice a day, then this product and quantity will prevent cancer.8 What would happen to the demand for cranberry juice? On the other hand, suppose that a television report stated that the botulism epidemic currently spreading in our region is caused by peanut products and products containing any peanut derivative from a very large manufacturing plant in the region. What would happen instantly to the demand for these products? In today’s dynamic market, such events happen frequently.
These fallacies severely affect the forecast of a single SKU (what item, where located, and when in time) and therefore provide a very poor base for determining the required stock level of that SKU. It is clear that another approach (instead of a better forecast) is needed in order to make this stocking decision.
The TOC Way—The Distribution/Replenishment Solution
The Theory of Constraints (TOC) analyzes the impact of the supply together with the demand to compute the right level of stocks throughout the supply chain, with the emphasis on the supply side. In an extreme case, where it is possible to respond instantly to demand, there is no need to rely on a forecast at all. While this situation is, of course, unattainable in almost all business environments, a step in this direction should be considered. In the case of keeping the right amount of stock in the supply chain, the TOC objective in responding to the three questions (what, where, and when) is to have very good availability of the items at all the consumption points (the end users). This objective is limited by the availability of cash and space, which means that it is impossible to keep high stocks of all items at all locations, even when obsolescence is not an issue. Not only that, but also as will be explained later in this chapter, keeping too high stocks of low demand SKUs will lower the total sales overall.
The