A demand forecast (also referred to as a usage rate) is a prediction of the amount of each product that will be sold, transferred, used in an assembly, or otherwise consumed in the future. It’s easy to see that inaccurate forecasts can cause major problems for your company. When forecasted demand is far below actual demand for a product, you risk running out of products and disappointing your customers. If forecasted demand exceeds actual usage, you’ll probably be faced with large quantities of dead stock and slow-moving inventory. For your company to be successful, your demand forecasts for products must be as accurate as possible.
Most software packages offer some demand forecasting tools. Most do a fairly good job of predicting future usage. But some of them include attributes that prevent them from doing the best possible job of predicting future demand in every situation. In this article we’ll look at one situation that few software packages handle well: items with long lead times. These may be products that are made to order or are imported from overseas.
When most packages recalculate purchasing parameters, they forecast demand for the upcoming month. For example, at the end of December, the system will forecast demand of products for January. Replenishment decisions are made based on this new forecast. This works well if your products have short lead times, say a week or ten days. But what if some of your products have extended lead times? For example, an item may have a twelve-week lead time. If you order the product at the end of December, you’ll receive the shipment sometime around April 1st. At the end of December, predicting how much you will sell in January doesn’t help the buyer at all. He or she needs to know what April’s demand for the product will be.
Fortunately, it’s fairly easy to consider extended lead times in forecasting the future demand of products. The process is best illustrated with an example:
- Add the current predicted lead time (expressed in days) for the product being forecast to the current date. The resulting date falls within the period for which to forecast demand. We’ll call this the “forecast period.” Depending on your system, the forecast period can be a week, a month, an entire season, or a certain number of days. In this example we’ll consider the forecast period to be one month.
- Once you determine the proper forecast period, you need to calculate a demand forecast for that period. This demand forecast should be determined by three elements:
- Historical usage.
- Changing trends.
- Known changes in demand that are not reflected in past history or trends.
Historical Usage
There are many formulas that use past usage to forecast demand. In fact, EIM now uses 29 different formulas to forecast demand customers’ finished goods inventory. Each formula is appropriate in different circumstances. As we discussed previously, a formula designed to forecast demand for the upcoming month is not appropriate for products with long lead times.
We’ve found that in most cases the best way to forecast usage for products with long lead times is to look at the usage surrounding the same forecast period, last year. For example, if we’re forecasting demand for April, 1999, we’ll determine an average usage per month by averaging the usage recorded in March, April, and May, 1998.
Trends
There is a problem with forecasting demand with history that is a year old: Usage of a particular item may have dramatically increased or decreased during the past 12 months. For this reason, a “trend factor” is applied to reflect the changes in your volume of business.
There are many ways to calculate a trend factor. An accurate trend factor can usually be calculated by comparing total usage (expressed in units) during the three months prior to the date you are calculating the forecast to the total usage recorded during same three months in the previous year. In our example we are calculating our forecast in December, 1998. So, we’ll compare the total usage in October, November, and December, 1997 to the total usage in October, November, and December, 1998. If the usage has increased by 10%, we’ll increase the average usage determined above by 10%. If total usage during the three-month period has decreased by 20% compared to the same period last year, we’ll decrease the average usage by 20%.
Known Changes in Future Demand
What if you know that, because of a new customer, usage of a specific item will increase by approximately 50 pieces per month? Or what if a new product is introduced and you predict it will replace about 50% of the sales of an existing stocked item? Neither of these circumstances is reflected by either usage history or trends. A mechanism must be provided for a buyer to manually enter these known changes in demand.
The demand forecast for April is determined by adding the calculated average usage (adjusted by a trend factor) to these known changes in future demand for the forecast period.
This month we’ve looked at forecasting items with long, relatively consistent lead times. It’s obvious that many of the traditional methods for forecasting lead times don’t apply to this situation. In future articles we’ll look at other “special” forecasting circumstances. For example:
- Setting up an import schedule designed to keep inventory levels to a minimum while avoiding stock outs.
- Buying for several months or an entire season.
- Dealing with long, irregular lead times.
- Replenishment orders that fill tractor-trailers, containers, or rail cars that don’t create dead stock and slow-moving inventory.
Our goal is to determine the best forecasting model for each stocked product, because the more accurate the forecast, the better you’ll be able to maximize both customer service and the return on your inventory investment. Please let us know if you have questions or comments.