Purchasing Recommendations produced by EPR systems will always be based on the assumption that the company does not want to run out of stock. This basic premise is flawed and really only applies when the product is in very high demand, it is difficult to get or there is no substitutable alternative.
A distributor makes a conscious decision to keep Inventory of certain products and to resell them to its customers. As each new product is taken on, certain attributes of the product are recorded:- static information such as Code, Description, Units of Measure, selling information such are Price, Discount Structures, Web Image etc.
A key attribute that is rarely considered in ‘Expected Service Level’. Consider the following:
The ‘Required Service Level’ is rarely considered when products are introduced into a company but it is fundamentally important in the control of Inventory. A Required Service Level of 80% tells the purchasing manager that he/she must ensure that the product can be supplied out of stock to a customer 4 times out of every 5 requests. It is OK to be out of stock!!
The Service Level is not something that should be randomly set. While certain assumptions have to be made for new products, each product’s history can contribute valuable information. Once accurate demand figures are in place, the analysis of this demand will result in a bell curve like this:
The dots represent the demand for a product at a particular point in time. Let’s imagine that each dot represents a week. We can easily calculate the average demand. This is the center point of the graph. We can see that some of our dots are far to the left, telling us that some weeks the demand is low, and on the other side, we have some weeks that the demand is high.
The Empirical Rule tells us that if we calculate the average demand plus or minus 1 standard deviation, then 68% of our samples will be in the section of the graph indicated by the pink lines. If we want to include 95% of the samples, then we need to calculate 2 standard deviations either side of the mean (denoted by the green bars). If we want to include 99.7% of the samples, then we need to calculate 3 standard deviations (denoted by the light blue bars).
Focusing on the 68% of the graph, we see that half of it (34%) is to the left of the average and the other half is to the right. When demand is higher than the average (the right-hand side of the graph), there are only 16% of the dots that are not covered.
We can apply this logic to our Service Level calculation. If we want to ensure that we have this item available 84% of the time, then we need to hold the average demand quantity plus 1 standard deviation. If the average demand is 100 and the standard deviation is 10, then our aim is to have 110 in stock at the start of each week.
By setting our target stock level at the average plus 2 standard deviations (e.g. 120), we can be 97.5% confident is being able to satisfy demand.
The required service level can be set by applying an ABC code to the product with ‘A’ indicating that 97.5 Service is required, B indicating that 85% service level is required and C indicating that 50% service level is required.
Different rules apply to different distribution curves. The Empirical Rule will not apply if the product is sold on an irregular basis. Therefore the calculation engine must be able to analyse the history of the product and identify whether the product can use the Empirical Rule qualified by the service level indicator.
Additional distribution analysis methods need to be examined in the detailed specification.
