Part I THE BIG PICTURE Sales Management Resources

Part I THE BIG PICTURE Sales Management Resources: Estimating Potentials and Forecasting Sales

WHY FORECAST? n n One of the keys to success in sales is knowing where customers are located and being able to predict how much they will buy. Firms have found that sales potential data are indispensable to developing a sales program, particularly in setting up territories, assigning quotas, developing budgets, and comparing sales performance of individual salespeople. Sales forecasting is so important that more than 50 percent of firms include this topic in their sales manager training programs. Inaccurate demand predictions can have disastrous effects on profitability.

WHAT IS MARKET POTENTIAL? n n Market potential is an estimate of maximum demand in a time period based on the number of potential users and their purchase rate. Actual industry sales are usually less than market potential, as shown in Figure SMR 2. 1 Actual sales are less than potential because it takes time to convince people to buy discretionary items such as digital video disc players and because some people can’t afford them. Company sales potential is a portion of total industry demand. It is the maximum amount a firm can sell in a time period under optimum conditions.

Figure SMR 2 -1 Relations Among Market Potential, Industry Sales, and Company Sales Market potential Industry forecast Basic Demand Gap Industry Sales Company potential Company forecast Company Demand Gap Actual Forecast 1 2 3 4 5 6 7 8 9 10 11 12 Custom time period

Estimating Potentials n n n n All estimates of potential are based on two key components— the number of possible users of the product the maximum expected purchase rate. Sometimes you can get estimates of these numbers from trade associations or commercial research associations, but you have to come up with your own potential figures, broken down by geographical area, industry and customer type. The initial approach for estimating the number of buyers is to use secondary sources. A wide variety of commercial data are available that provide the potential number of buyers, size of firms, age of consumers, income levels, and locations. Purchase rates are usually derived from trade organizations or government publications. For existing products, you can use the ratio of current sales to the number of households or sales person. These ratios can be obtained from trade publications such as those from the Conference Board, or they can be calculated from published data.

Buying Power Index Method n n Market potentials for consumer goods are usually estimated by constructing indexes from basic economic data. Perhaps the most popular multifactor index of area demand is the Buying Power Index (BPI), published each year by Sales & Marketing Management magazine. This index combines estimates of population, income, and retail sales to give a composite indicator of consumer demand in 922 geographic areas known as Core Based Statistical Areas (CSBSAs). Buying Power Index values are used to help managers allocate selling efforts across geographic regions.

Table SMR 2 -1 Data Used to Calculate Buying Power Index 2002 Effective Buying Income Amount ($000, 000) Total United $5, 466, 880 States Atlanta Metro $ 99, 691 2002 Total Retail Sales Percentage Amount of United States ($000, 000) 2002 Estimated Total Population Percentage Amount of United States ($000, 000) Percentage of United States Buying Power Index 100. 00% $3, 906, 482 100. 0% 292. 936 100. 0% 100. 00 1. 824% $ 69, 071 1. 768% 4. 704 1. 606% 1. 7636

NAICS Method for Business Markets n Business market potential can be built up from data made available through the U. S. Census of Manufacturers. The Census of Manufacturers, which is available every 5 years, combines businesses into North American Industry Classification System (NAICS) codes according to products produced or operations performed.

Table SMR 2 -2 Estimating the Market Potential for Food Machinery in North Carolina (2) Number of Machines Used per 1000 Workersb Market Potential (1 x 2) NAIC Code Industry (1) Production Employeesa 3112 Grain Milling 811 24 19. 5 3122 Tobacco Mfg. 9, 328 15 139. 9 3121 Beverages 1, 757 3 5. 3 164. 7 a The production employee data are from the 1997 Economic Census of Manufacturing, Geographic Area Series, North Carolina, p. NC 8. The codes are the new NAIC codes b Estimated by manufacturer from past sales data.

QUALITATIVE SALES FORECASTING n n Sales forecasting is concerned with predicting future levels of demand. These projections are vital for budgeting and planning purposes. For new products, a few simple routines can be employed. The absence of past sales means that you have to be more creative in coming up with predictions of the future. Sales forecasts for new products are often based on executive judgments, sales force projections, surveys, and market tests. We will begin our discussion of forecasting techniques by focusing on subjective methods that are based on interpretations of business conditions by executives and salespeople.

Sales Force Composite n A favorite forecasting technique for new and existing products is the sales force composite method. With this procedure, salespeople project volume for customers in their own territory, and the estimates are aggregated and reviewed at higher management levels. The territory estimate is often derived based on demand estimates for each of the largest customers in the territory, the remainder of the customers as a group, and then for new prospects.

Jury of Executive Opinion n This technique involves soliciting the judgment of a group of experienced managers to give sales estimates for proposed and current products. The main advantages of this method are that it is fast and it allows the inclusion of many subjective factors such as competition, economic climate, weather, and union activity.

Leading Indicators n n n Where sales are influenced by basic changes in the economy, leading indicators can be a useful guide in preparing sales forecasts. The idea is to find a factor series that is closely related to company sales, yet for which statistics are available several months in advance. Changes in the factor can then be used to predict sales directly, or the factor can be combined with other variables in a forecasting model. Some of the more useful leading indicators include prices of common stocks, new orders for durable goods, new building permits. Leading indicators are sensitive to changes in the business environment and they often signal turns in the economy months before they actually occur.

When Should Qualitative Forecasting Methods Be Used? n n Qualitative methods are often used when you have little numerical data to incorporate into your forecasts. New products are a classic example of limited information, and qualitative methods are frequently employed to predict sales revenues for these items. Qualitative methods are also recommended for those situations where managers or the sales force are particularly adept at predicting sales revenues. In addition, qualitative forecasting methods are often utilized when markets have been disrupted by strikes, wars, natural disasters, recessions, or inflation.

When Should Qualitative Forecasting Methods Be Used? n Under these conditions, historical data are useless, and judgmental procedures that account for the factors causing market shocks are usually more accurate. Managers should calculate and record the forecasting errors produced by the qualitative techniques they employ so that they will know when these methods are best employed.

QUANTITATIVE SALES FORECASTING n n n Seasonal Adjustments Before we discuss data-based forecasting techniques, it’s important to understand how seasonal factors influence predictions of the future. Sales forecasts are often prepared monthly or quarterly, and seasonal factors are frequently responsible for many of the shortrun. When historical sales figures are used in forecasting, the accuracy of predictions can often be improved by making adjustments to eliminate seasonal effects.

Table SMR 2 -4 Calculating a Seasonal Index from Historical Sales Data Quarter 1 2 3 4 Four-Year Quarterly Average 1 49 57 53 73 58. 0 0. 73 a 2 77 98 85 100 90. 0 1. 13 3 90 89 92 98 92. 3 1. 16 4 79 62 88 78 76. 8 0. 97 Four year sales of 1268/16 = 79. 25 average quarterly sales a Seasonal index is 58. 0/9. 25 = 0. 73 Seasonal Index

Seasonal Adjustments n n n A seasonal index of 0. 73 is obtained. This number indicates that seasonal factors typically lower firstquarter sales by 27 percent. Actual sales, such as those shown in Table SMR 2 -4, are simply divided by the appropriate index numbers to give a set of depersonalized data. Sales forecast are then prepared using the deseasonalized sales figures. For example the deseasonalized sales data for the four quarters of the first year in Table SMR 2 -4 would be 67, 68, 78, and 81 for quarters one, two, three, and four, respectively. 1. Seasonal adjustments are widely used in business. 2. Seasonal adjustments reduce forecasting errors.

Naive Forecasts n n Time series forecasts rely on past data to provide a basis for making projections about the future. The naive forecast is the simplest numerical forecasting technique and is often used as a standard for comparison with other procedures. This method assumes that nothing is going to change and that the best estimate for the future is the current level of sales. For example, actual sales of 49 units observed in quarter 1 in Table SMR 2 -4 can be used to predict sales in quarter 2. Naive forecasts for the last three quarters of year 1 would be

Naive Forecasts

Naive Forecasts n n The naive approach may also be used with deseasonalized sales figures, such as those calculated in the previous section. Recall that the seasonally adjusted sales figure for the first quarter of year 1 was 67, so the naive forecast sales in the second quarter would also be 67. Seasonally adjusted, the forecast for the second quarter would be 76 (67 × 1. 13 = 75. 7).

MAPE n n n In order to compare forecasting accuracy across several time periods, most forecasting professionals use the mean absolute percentage error (MAPE) method. 5 The formula for calculating MAPE is: Where n is the number of periods for which forecasts are to be made. MAPE calculates the percentage forecasting error for each period without regard to whether the errors are positive or negative, adds up the errors, and divides by the number of periods being forecast. The main advantage of MAPE is that it allows easy comparison of forecasting errors across product categories and companies.

MEAN ABSOLUTE PERCENTAGE ERROR (MAPE)

Trend Projections n n The use of trends to project sales is a popular technique among business firms. With this method, the analyst estimates trends from past data and adds this figure to current sales to obtain a forecast. For example, in Figure SMR 2 -2 sales increased from 10 units in period 2 to 20 units in period 3, suggesting a trend of 10 units period. A unit rate of change forecast for period 4 would combine current sales of 20 plus 10 units of trend for a total of 30. Trends can also be expressed as a percentage rate of change. Note that the percentage rate of change method and the unit rate of change procedure give different sales forecasts.

Figure SMR 2 -2 Comparing Trend Forecasting Methods 50 40 Sales Percent rate of change forecast 30 Unit rate of change forecast 20 Naïve forecast Moving average forecast 10 0 1 2 3 4 Time Period 5

Moving Averages With the moving average method, the average revenue achieved in several recent periods is used as a prediction of sales in the next period. n The formula takes the form :

CALCULATING A MOVING AVERAGE FORECAST where Ft+1 St n = forecast for the next period = sales in the current period = number of periods in the moving average

Moving Averages n n A crucial issue in using moving averages is determining the ideal number of periods (n) to include in the average. With a large number of periods, forecasts tend to react slowly, whereas a low value of n leads to predictions that respond more quickly to changes in a series. The optimum number of periods can be estimated by trial and error or with computer programs. A characteristic of moving averages that distracts from their ability to follow trends is that all time periods are weighted equally. This means that the oldest and most recent periods are treated the same in making up a forecast. A popular technique that overcomes this problem is exponential smoothing.

MOVING AVERAGE FORECASTING EXAMPLE Quarter Actual sales Two-period moving average 1 2 3 4 49 77 90 79 63 83. 5

FORECASTING WITH MOVING AVERAGES 1 2 Time Periods 3 Actual sales Seasonally adjusted sales Two-period moving average forecast seasonally corrected Three-period moving average forecast seasonally corrected Two-period moving average forecast 49 67 77 68 90 78 79 81 57 78 98 87 78. 3 70. 1 58. 0 89. 8 68. 9 55. 2 Three-period moving average forecast 89. 3 F 3 = ( S 1 + S 2 ) x I 3 2 F 4 = ( S 1 + S 2 + S 3 ) x I 4 3 = ( 67 + 68 ) x 1. 16 2 = 78. 3 4 5 6 = ( 67 + 68 + 78 ) x 0. 97 3 = 68. 9

Exponential Smoothing An important feature of exponential smoothing is its ability to emphasize recent information and systematically discount old information. A simple exponentially smoothed forecast can be derived using the formula:

CALCULATING AN EXPONENTIAL SMOOTHING FORECAST where = smoothed sales forecast for period t and the forecast for period t + 1 α = the smoothing constant St = actual sales in period t -1 = smoothed forecast for period t – 1

EXPONENTIAL SMOOTHING FORECASTING EXAMPLE Quarter Actual sales 1 2 3 4 49 77 90 79 60. 2 72. 1 Smoothed forecast

Exponential Smoothing n n n The major decision with exponential forecasting is selecting an appropriate value for the smoothing constant (á). Smoothing factors can range in value from 0 to 1, with low values providing stability and high values allowing a more rapid response to sales changes.

Time Series Regression n n In time series regression, the relationship between sales (Y) and a period of time (e. g. , week, month, quarter, or year) (X) can be represented by a straight line. The equation for this line is Y = a + b. X, where a is the intercept and b shows the impact of the independent variable on Y. The key step in deriving linear regression equations is finding values for the coefficients (a, b) that give the line that best fits the data.

Figure SMR 2 -3 Fitting a Trend Regression to Seasonally Adjusted Sales Data 90 80 3. 6 Sales 70 Y = 63. 9 + 3. 5 X 63. 9 60 50 0 1 2 3 Time Period 4 5 6

Turning Points n n n At several points in this discussion we have mentioned the idea of turning points. A turning point is a sudden change in a trend. For instance, a decline in sales after several years of moderate growth would be considered a turning point, if sales continued to decline in subsequent years. The numerical forecasting methods we have discussed make projections from historical data, and most of them do a poor job of predicting turning points in a time series. Percentage rate of change, unit rate of change, and time series regression are all notoriously poor predictors of series that change direction. Naive, moving average, and exponential smoothing are somewhat better because they tend to lag and then adapt to new information.

When Should Quantitative Forecasting Methods Be Used? n n Quantitative forecasting techniques are best employed when you have access to historical data. It is also helpful if the time series you are trying to forecast are stable and do not frequently change direction. Quantitative methods have distinct advantages in situations where you must make frequent forecasts for hundreds or thousands of products. Because of the large number of calculations required by quantitative forecasting procedures, analysts need access to computers and appropriate forecasting software. .