Judgment and Survey Techniques

At some level, every forecast is ultimately someone’s judgment. Some techniques, though, rely more on people’s opinions or estimates and are called judgment techniquesForecasting methods that rely on someone’s estimate(s).. Judgment techniques can include customer (or channel member or supplier) surveys, executive or expert opinions, surveys of customers’ (or channel members’) intentions or estimates, and estimates by salespeople.

Time Series Techniques

Time series techniques examine sales patterns in the past in order to predict sales in the future. For example, with a trend analysisA group of forecasting methods that base the future period of sales (or another variable) on the rate of change for previous periods of time., the marketing executive identifies the rate at which a company’s sales have grown in the past and uses that rate to estimate future sales. For example, if sales have grown 3 percent per year over the past five years, trend analysis would assume a similar 3 percent growth rate next year.

A simple form of analysis such as this can be useful if a market is stable. The problem is that many markets are not stable. A rapid change in any one of a market’s dynamics is likely to result in wide swings in growth rates. Just think about auto sales before, during, and after the government’s Cash for Clunkers program. What sold the previous month could not account for the effects of the program. Consequently, if an executive were to have estimated auto sales based on the rate of change for the previous period, the estimate would have been way off.

Figure 16.10

The federal government’s Cash for Clunkers program resulted in a significant short-term increase in new car sales and filled junkyards with thousands of clunkers!

Source: Wikimedia Commons.

The Cash for Clunkers program was an unusual situation; many products may have wide variations in demand for other reasons. Trend analysis can still be useful in these situations but adjustments have to be made to account for the swings in rates of change. Two common adjustments are the moving averageA trend analysis type of forecasting method that estimates sales (or other variable) based on an average rate of change over a group of previous periods of time; the rate changes (moves) as the oldest period is dropped off and the most recent period added in., whereby the rate of change for the past few periods is averaged, and exponential smoothingA method of trend analysis forecasting that weights more recent periods of time more heavily than more distant periods of time., a type of moving average that puts more emphasis on the most recent period.

Correlates and Other Models

A number of more sophisticated models can be useful in forecasting sales. One fairly common method is a correlational analysisA form of trend analysis that estimates sales based on the trends of other variables., which is a form of trend analysis that estimates sales based on the trends of other variables. For example, furniture-company executives know that new housing starts (the number of new houses that are begun to be built in a period) predict furniture sales in the near future because new houses tend to get filled up with new furniture. Such a correlate is considered a leading indicatorA correlate that occurs before the variable being forecasted (e.g., permits to build new houses is a leading indicator of building material sales because permits are issued before the materials are purchased)., because it leads, or precedes, sales. The Conference Board publishes an Index of Leading Indicators, which is a single number that represents a composite of commonly used leading indicators. Some of these leading indicators are housing starts, wholesale orders, orders for durable goods (items like refrigerators, air conditioning systems, and other long-lasting consumer products), and even consumer sentiment, or how consumers think the economy is doing.

Market Tests

A market testThe test launch of a product’s complete marketing plan to ensure that it reaches buyers, gets positive reactions, and generates sales of the product. is an experiment in which the company launches a new offering in a limited market in order to gain real-world knowledge of how the market will react to the product. Since there isn’t any historical data on how the product has done, response models and time-series techniques are not effective. A market test provides some measure of sales in response to the marketing plan, so in that regard, it is like a response model, just based on limited data. The demand for the product can then be extrapolated to the full market. However, remember that market tests are visible to your competitors, and they can undertake actions, such as drastic price cuts, to skew your results.

Figure 16.11

HEB uses Waco, Texas, as a test market, combining data from its loyalty program with sales data to see who buys what and at what price.

Source: Wikimedia Commons.

The grocery chain HEB uses Waco, Texas, as a test site. HEB has a loyalty program that enables it to collect lots of data on its customers. When HEB wants to test market a new product, the firm does it in Waco, where individual customer data can be combined with sales data. Testing in Waco tells HEB who is likely to buy the product and at what price, information that makes extrapolating to their larger market more accurate.

Pick the Right Method(s) for Your Business and Your Decision

Some products have very short selling cycles; others take a long time to produce and sell. What is appropriate for a fast-moving consumer good like toothpaste is not appropriate for a durable good like a refrigerator. A response model might work for Crest toothpaste in the short term, but longer-term forecasts might require a sophisticated time-series technique. By contrast, Whirlpool might find, for example, that channel surveys are better predictors of refrigerator sales over the long term.

Use Multiple Methods

Since forecasts are estimates, the more estimates generated from various methods, the better. For example, combining expert opinions with a trend analysis could help you understand not only what is happening but also why. Every forecast results in decisions, such as the decision to hire more people, add manufacturing capacity, order supplies, and so forth. In addition, practice makes perfect, as they say. The more forecasts you have to make and resulting decisions you have to live with, the better you will get at forecasting.

Use Many Variables

Forecasting for smaller business units first can result in greater accuracy. For example, JCPenney may estimate sales by region first, and then roll that information up into a national sales forecast. By forecasting locally, more variables can be considered, and with more variables comes more information, which should help the accuracy of the company’s overall sales forecast. Similarly, JCPenney may estimate sales by market segment, such as women over age fifty. Again, forecasting in a smaller segment or business unit can then enable the company to compare such forecasts to forecasts by product line and gain greater accuracy overall.

Use Scenario-Based Forecasts

One forecast is not enough. Consider what will happen if conditions change. For example, how might your forecast change if your competitors react strongly to your strategy? How might it change if they don’t react at all? Or if the government changes a policy that makes your product tax free? All of these factors will influence sales, so the smart executive considers multiple scenarios. While the executive may not expect the government to make something tax free, scenarios can be created that consider favorable government regulation, stable regulation, and negative regulation, just as one can consider light competitive reaction, moderate reaction, or strong reaction.

Track Actual Results and Adjust

As time goes on, forecasts that have been made should be adjusted to reflect reality. For example, Katie Scallan-Sarantakes may have to do an annual forecast for Scion sales, but as each month goes by, she has hard sales data with which to adjust future forecasts. Further, she knows how strongly competition has reacted and can adjust her estimates accordingly. So, even though she may have an annual forecast, the forecast changes regularly based on how well the company is doing.