Monitoring Your Forecast Accuracy

Replenishment capabilities in most computer software packages answer two vital questions:

 

  • When a stocked item needs to be reordered to avoid a stockout?
  • How much of the product should be ordered?

 

A critical factor in answering these questions is an accurate prediction of what will be sold, transferred or otherwise consumed in the future.  This estimate of future demand is a forecast.

 

Without an accurate forecast your system can’t effectively answer those two questions.  After all, wouldn’t your buying decisions be different if you sold 10 pieces or 1,000 pieces of an item in a typical month?  Even what seems to be a small difference in forecast quantities can make a big difference in your replenishment plans.  For example, if you forecast selling one piece per day and had a seven day lead time you would have to reorder the product when you had no less than seven pieces in stock.  However, if your forecast was two pieces per day, a replenishment order would have to be ordered when there were no less than 14 pieces in your warehouse.

 

Do you monitor the accuracy of your forecasts?  In a study involving 78 organizations we performed several years ago we found that the median average forecast error for individual products was 381%.  That means that forecasts were almost four times greater than or less than actual usage!  But we also noticed that organizations in the top 10% in their industry (based on profitability) had an average forecast error that was less than 10% of the average forecast error of all organizations in their industry.  It is obvious that more accurate predictions of future usage allow you to make better stocking decisions.

 

To improve the accuracy of your forecasts:

  1. calculate the current forecast error for each of your stocked items, last month, using the formula:

 

Absolute Value of (Actual Usage – Forecast) ÷ Lower of Forecast or Actual Usage

 

If you have a forecast for a product of 50 pieces and experience actual usage of 75 pieces, the forecast error would be 50% [Absolute Value of (75 – 50) ÷ 50 = 50%].  The forecast error would be the same if these number were reversed.  That is a forecast of 75 pieces and actual usage of 50 pieces [(Absolute Value of (50 – 75) ÷ 50 = 50%].

 

  1. Starting with your “A” ranked products, examine each item with a high forecast error (maybe greater than 50%). Did the item experience a stockout or unusual sales activity that could not have been predicted?  If not, we need to find a better way to forecast future demand of the product.

 

Keep in mind that different items in your inventory, even items in the same product line, will experience very different patterns of usage.  As a result, they will require different forecasting formulas.

 

Next month we will start to explore how to determine the best way to forecast each individual item to help you achieve the goal of effective inventory management.