It is no secret that an accurate forecast of the future demand of a product is crucial in achieving the four “rights” of effective inventory management: that is, getting the right quantity of the right item to the right location at the right time. As we’ve discussed in previous articles, products with different patterns of usage require different forecasting methods. The forecast for items with recurring usage is usually based on four elements:
- Some sort of average of past usage.
- A trend derived from past usage.
- Future anticipated usage that is not revealed in past usage or trends.
- A forecast horizon reflecting when material ordered today can be received and the length of time for which inventory must be purchased.
If an item has recurring usage (that is, it is sold or used on a regular basis) we can test various formulas that apply different factors to each of the four elements to determine the best method of forecasting future demand of each item. But applying these elements to an item with sporadic activity (i.e. one that is not sold on a regular basis) produces strange results. Look at the usage history of this item:
This product is not sold that often, but when it is sold the customer wants 12 pieces. Any forecast formula that includes some sort of average of past usage will probably base the product’s replenishment parameters on an average usage per month of about three pieces (36 pieces ÷ 12 months). Will stocking the product based on selling an average of three pieces per month satisfy our customer? Probably not. When they order the product they request 12 pieces. This typical purchase quantity should serve as the basis of our minimum and maximum quantities. Maybe we will reorder 12 pieces of the item when there are 12 left on the shelf. Or maybe we are willing to risk a stock out and will wait until the stock of the item is completely depleted before we order 12 more pieces.
Because different methods are used to control the replenishment of products with recurring and sporadic usage, it is imperative that we are able to accurately and consistently identify whether each stocked product falls into one category or the other.
Sporadic usage items are sold infrequently, maybe in less than six of the past 12 months. Can we say that any item with usage in less than six of the past 12 months has sporadic activity? The item in the above example meets this criterion, but so does this one:
The product is only sold in five months out of the year (June through October). But does it have sporadic sales? No. The product appears to have recurring sales during a specific season of the year. If the average sale quantity of the product was one piece, and we based our replenishment parameters on this “normal” transaction, our company would probably experience significant stocking problems. This item is probably best forecast using a seasonal formula that includes all four elements for forecasting items with recurring usage.
But how do you differentiate between items with sporadic sales and seasonal items? Retired industry “guru” Gordon Graham in his book, Distributor Survival in the 21st Century, suggests that an item is usually seasonal if 80% or more of the sales in the previous 12 months occurred in just a six-month time period*. The item in the last example meets Gordon’s seasonal criteria, but look at the usage of this next item:
Four of the five pieces (i.e. 80% of the total) were sold in a six-month period. But this appears to be a product that experiences sporadic, not seasonal usage. Graham recognized this problem. He suggested that a buyer manually review each item selected by his 80% rule and determine whether it has seasonal or sporadic activity. For many companies that means sifting through thousands of product history records. That’s a lot of work – and there is a strong possibility that some items will be misidentified or overlooked.
Other forecasting methods differentiate between seasonal items and those with sporadic sales by examining the total usage in a twelve-month period. Items with sporadic sales should have low total usage (say, less than 12 pieces per year). Right? Well, what if you sold a large quantity just two or three times a year:
Should the purchasing parameters of the above item be based on an average quantity sold per month of 100 pieces (1200 pieces ÷ 12) or the normal usage quantity of 600 pieces? The “Total Quantity” maxim doesn’t reliably differentiate between items with a high volume of sales in a limited number of transactions and those that have recurring usage activity.
We think we’ve found a better way to identify sporadic items. Actually, it is a simple, reliable way to identify items that do not have sporadic activity! We start by dividing the past 12 months into nine four-month groups:
If a product does not have usage in at least three of the four months or at least one of the nine four-month groups, it is identified as having sporadic usage. Let’s take another look at the item in the last example:
Notice that none of the nine groups has usage in three or four months – therefore it is correctly identified as having sporadic usage. Let’s see if this method of identifying seasonal items will work with other items we’ve examined:
Example #1:
Example #2:
Once we’ve determined that an item does not have sporadic usage, we can test several formulas (both seasonal and non-seasonal) to determine the best forecasting method for that particular item. Minimum and maximum quantities for the sporadic usage items will be based on the normal sales quantity.
It’s easy to see that forecasting future usage of an item using an incorrect formula will result in stocking the wrong quantity of the wrong item in the wrong location at the wrong time. To achieve effective inventory management, it is essential to be able to differentiate between items with sporadic sales and those with recurring usage activity.
*Graham, Gordon, Distributor Survival in the 21st Century, Inventory Management Press 1992, page 40.