Smoothing Model Helps Predict Sales: Forecasting the Future
Gross, Kevin D., Journal of Property Management
A critical aspect of managing property in today's real estate marketplace is planning for the future. Increasingly, owners and investors are requiring real estate managers to perform long-term forecasting functions. For example, in order to effectively project a property's income, managers are being asked to anticipate monthly, quarterly, and/or annual estimates of sales of tenants which have percentage leases. Accuracy in forecasting such information is essential as poor estimates may result in negative variances, thus reducing capital for expenditures and anticipated dividends to owners.
One effective forecasting technique to consider is the exponential smoothing model. This form of forecasting utilizes a weighted average of past time series values in order to predict the value of the time series in the succeeding period. Smoothing methods are easy to use and generally provide a high level of accuracy for short-range forecasts such as a forecast for the following time period. Exponential smoothing uses weighted moving averages in which only one weight-the weight for the most recent observation-is selected. The weight for the other data values is automatically computed and becomes increasingly smaller as the observations move further into the past.
The exponential smoothing model can best be understood by viewing the actual program for this forecasting technique. (To download a free copy of this Excel program, visit: http://www.irem.org/jpm651.htm and save on your hard drive.) The remainder of this article will discuss the step-by-step process of utilizing this forecasting program.
Using the Forecasting Model Program
Utilizing the exponential smoothing model program to forecast quarterly, monthly and annual estimates is a simple process. Simply perform the following steps. (Note:
The following explanation of the model is based on a "quarterly" forecast, but the process is the same for the monthly and annual segments of the program.)
1. Open the program and click on the tab titled "Quarterly."
2. Click on the yellow cell titled "Enter Last Quarter" and a drop-down menu will appear. Scroll down this menu until you locate the quarter for which you have the most recent data. (Please note that you may only input data in the bright yellow cells.)
3. Next, in cells B-11 through B-30, enter the actual sales or data that corresponds with the quarter listed to the left of the yellow cells. If you do not possess all sales data for all quarters, enter the data that you do have. Because this program is set, you will have to make adjustments if you do not have an entire history of data. To compensate for this, simply enter the same data in your blank cells that you have listed in your oldest known period. For example, if you have three years of quarterly data (12 quarters), there will be two years (8 quarters) which effectively will be blank and which will have zeros (0) in them. Do not leave these cells empty. Instead, insert the oldest actual income into all blank cells.
4. Move your curser to the cell titled, "Smoothing Constant." When the smoothing constant equals one hundred (100%), the model is said to be "naive." The problem with this is that it responds only to what happened in the preceding period. This means it overreacts to what might have been simply random variation. On the other hand, when the smoothing constant equals zero (0%), the forecast wholly ignores what happened during the preceding period and continues to make the same forecast as before. Neither extreme is realistic. Selecting a suitable value for the smoothing constant amounts to finding a middle ground between overreacting and underreacting to what happened during the preceding period.
When demand varies widely above and below an average value with little or no overall trend, the values of the smoothing constant should be small. …