Forecasting Methods In Big-Box Retailing

While considered a form of financial voodoo in many industries, accurate forecasting is vitally important in any industry that needs to make business decisions based on what the future holds. Forecasting demand is important to manufactures when determining how much of a product to produce, and equally as important to industries such as retailers when trying to predict how much product the demand bears. There are several methods of forecasting commonly used, with the choice primarily being guided by the demand one is trying to predict. The following are a few methods used with examples of how certain industries use them as well as how these methods are used in my current industry; home improvement retailing.

The first method is historical analogy. This is a qualitative method of forecasting meaning that it is fairly subjective and based on estimates and opinions. A common use for this method is when a firm is trying to forecast demand for a new product (Chase, 2006). A company could use past demand of a similar product to help predict future demand for the new product. Chase, et al used an example of a firm that produces toasters who wants to carry a coffee maker. They could reasonably use the toaster history as a possible growth model. While the two appliances have very different purposes, their similarity in other aspects is alike enough to make this method viable. For instance, they are both small countertop appliances. Their price points are relatively similar. A specific demographic with a demand for a toaster may have a similar demand for a coffee maker.

Seasonal forecasting is simply a time series method of forecasting that capitalizes on a seasonal component of demand. The component variation can be either additive or multiplicative. Differing from an historical analogy method of forecasting, a time series is used to predict future demand based on past data (Chase, 2006). This lends itself as a better choice for estimating demand of products with a long enough history to be relevant.

Noting that a time series can be defined as "chronologically ordered data that may contain one or more components of demand" including trend and seasonal demand, one can base their forecast on both components concurrently (Chase, 2006). The differences are in how these components relate to each other. An additive variation assumes that the variation is independent of the trend. Imagine a plot of a retail store's sales with the dollar amount on the y-axis and the time (in months or years) along the x-axis. If the trend variation was on an upward slope due to population growth in the area, seasonal variances would be independent of this trend. Seasonal variations such as colder-than-normal winters or wetter-than-normal springs would be additive to the trend line.

In a multiplicative variation, the seasonal variance and the trend line are dependent on each other. This factor (or index) could be defined as "the amount of correction needed in a time series to adjust for the season." For a seasonal business, this index could be calculated by the ratio of the demand in a quarter over the average of all quarters. This could also be a monthly index and in a very seasonal business, this could range anywhere from 0.7 to 1.3. The following is a hypothetical example of sales of core shrubs in a nursery.

Sales ($) Flat Avg ($) Index

Jan 7,000 10,000 0.7

Feb 8,000 10,000 0.8

Mar 14,000 10,000 1.4

Apr 15,000 10,000 1.5

May 12,000 10,000 1.2

Jun 9,500 10,000 0.95

Jul 8,000 10,000 0.8

Aug 8,500 10,000 0.85

Sep 11,500 10,000 1.15

Oct 11,000 10,000 1.1

Nov 8,000 10,000 0.8

Dec 7,500 10,000 0.75

Total 120,000

The index in this sample case is calculated by dividing the actual sales by the average for the specified periodicity. While this is a snapshot of one particular year, by calculating the Seasonal Index from several years of past data, the index becomes a more accurate tool for forecasting. The following is a plot of the same data to simply highlight the seasonality of this hypothetical shrub business.

When preparing forecasts for data such as sales, it is worth noting that retailers must distinguish between what merchandise in their mix is truly dependent on these seasonality factors (Berman, 2004). Basically, they must understand what is staple merchandise and what is seasonal. For a common grocer, milk and bread are staples and watermelons are quite seasonal. As one may predict, a business that is significantly impacted by the weather (such as a nursery) could greatly benefit their forecasting efforts by using a seasonality and time series method.

Another method of forecasting is using leading indicators. This is a causal method of predicting demand based on the trend of another metric that happens to move in the same