Sales forecasts are common and essential tools used for business planning, marketing, and general management decision making. A sales forecast is a projection of the expected customer demand for products or services at a specific company, for a specific time horizon, and with certain underlying assumptions.
A separate but related projection is the market forecast, which is an attempt to gauge the size of the entire market for a certain class of goods or services from all companies serving that market. Sales and market forecasts are often prepared using different methods and for different purposes, but sales forecasts in particular are often dependent at least somewhat on market forecasts. Although the focus of this discussion will be on sales forecasting, a brief summary of market forecasting will help provide context.
A special term in studying sales and market forecasts is the word "potential." This refers to the highest possible level of purchasing, whether at the company level or at the industry or market level. In practice, full potential is almost never reached, so actual sales are typically somewhat less than potential. Hence, forecasts of potential must be distinguished from forecasts that attempt to predict sales realized.
Assessing market potential involves observing and quantifying relationships among different social and economic factors that affect purchasing behaviors. Analysts at the industry level look for causal factors that, when linked together, explain changes (upward or downward) in demand for a given set of products or services. This may be done on the local level, the national level, or even the international level. The economic and social variables that are deemed most important—those that historically have shown the most influence on demand—are then incorporated into some type of formula or mathematical model that attempts to predict future purchasing activity based on expected changes in the causal factors.
The simplest example would be to consider the influence of widely observed macroeconomic indicators such as gross domestic product (GDP) and employment rates. A simplistic model of market growth might indicate that based on time-series data from the past decade the restaurant market tends to grow at one and one-tenth times the rate of GDP when the national unemployment rate is less than 7 percent, and at four-fifths of the GDP growth rate when unemployment is greater than 7 percent.
Suppose an analyst wishes to create a two-year forecast for the national restaurant business. Using published estimates from government or private sector economists, the analyst might learn that next year's GDP is expected to grow at 2.9 percent and unemployment is expected to register at 6.7 percent. The following year, however, GDP growth is expected to slow to 1.9 percent and unemployment is expected to rise to 7.6 percent. Using the simple model outlined, the forecast for next year's restaurant sales growth would be based on the first condition observed, namely that market growth is somewhat (10 percent) higher than GDP growth when unemployment is relatively low. In other words, the first year's forecast would be 1.1 × 2.9, or 3.19 percent restaurant market growth. In the second year, the second condition would come into play—market growth is slower than GDP growth—since unemployment is expected to surpass 7 percent. Thus the forecast would be 0.8 X 1.9 percent, or 1.52 percent growth in demand for restaurant services.
While this example illustrates the basic process of forecasting, serious market forecasts would of course consider many more factors than GDP and unemployment. For instance, more sophisticated models might look at the changing demographics of the customer base (size, average income, and other attributes), the rate of inflation, changes in interest rates, and changes in related markets that could affect the market under consideration. Consequently, the formulas for obtaining market forecasts are considerably more complex. But, as in this example, many market forecasts do rely on economic or demographic data from government or other sources; the forecaster often doesn't need to come up with from scratch his or her own projections for, say, GDP and population growth. Many market forecasts also rely on published indexes, ratios, and averages for various economic and social factors that have been compiled in databases or in reference books.
Sales forecasting is an attempt to predict what share of the market potential identified in a market forecast a particular company expects to have. For very small companies that serve only a fraction of the total market, the company forecast may not even explicitly consider the market forecast or share, although implicitly, of course, the company's sales are subsumed under the total market size. In the other extreme, a monopoly's sales forecast is essentially the same as the market forecast.
Forecasts of different kinds are often prepared at different levels of a corporate enterprise. Managers of different stripes use forecasts for a variety of purposes, including marketing planning, resource\investment allocation, production scheduling, and labor recruitment. In some cases the uses are simply informational, but in many cases forecasts are the basis for major decisions like:
Yet sales forecasts are conditional in that they are only estimations and are highly interdependent with corporate strategy and actions. Some forecasts are developed before strategies and action plans are formulated; others are created to gauge the anticipated effects of an existing strategy.
A sales forecast may cause management to adjust some of its assumptions or decisions about production and marketing if the forecast indicates that (1) the current production capacity is grossly inadequate or excessive and (2) sales and marketing efforts are inconsistent with the expected outcomes. Management therefore has the opportunity to examine a series of alternate plans for changes in resource commitments (such as plant capacity, promotional programs, and market activities), changes in prices, or changes in production scheduling. Indeed, when a company is evaluating different courses of action it may develop separate forecasts for each option in order to assess the implications of each.
As a starting point, management analyzes previous sales experience by product lines, territories, classes of customers, or other relevant categories. This analysis is often on a detailed level, such as on a month-by-month, quarterly, or seasonal basis, in addition to looking at overall annual trends. Such detailed views will allow management to look for seasonality in new forecasts or even to devise strategies to improve sales during slow seasons.
Management needs to consider a time line long enough to detect significant patterns in its sales history. This period is typically five to ten years. If the company's experience with a particular product class is shorter, management might also examine discernible experiences of similar companies. The longer the view, the better management is able to detect patterns that follow cycles. Patterns that repeat themselves, no matter how erratically, are considered "normal," while variations from these patterns are "deviant." Some of these deviations may have resulted from temporary or fluke conditions, such as bad weather or uncommon events. Depending on the circumstances, figures may need to be normalized to remove the influence of such factors.
Forecasting may also consider how the company rates against its competitors in terms of market share, research and development, quality, pricing and sales financing policies, and overall public image. In addition, forecasters may evaluate the quality and size of the customer base to determine brand loyalty, response to promotions, economic viability, and credit worthiness.
If prices for products have changed significantly over the years, changes in dollar volume of sales may not correlate well with unit sales. To adjust for such discrepancies, a price index may be developed showing the relative prices of goods for a given year versus some reference year. Perhaps the simplest case would be if the company's prices moved exactly at the rate of inflation, in which case it could use the historical tables from the Consumer Price Index or Producer Price Index, depending on what market it serves, to adjust its figures. Using this information, the company can establish more stable projections that are not unduly skewed by price fluctuations or inflation trends.
The condition of the overall economy often influences the rate of growth (or decline) for particular markets and firms. Sales forecasters may consider any number of macroeconomic trends that have been shown to correlate with company sales, including GDP and inflation. General indicators like these can be essential in interpreting a sales forecast or recent sales history, as they will show, for example, whether the company's dollar sales are rising faster than the rate of inflation or whether the company is growing more rapidly than the economy on average.
Similarly, the company may consider its performance relative to its industry, the secular trend. While the secular trend represents the average for the industry, it may not be "normal" for a particular company. A comparison of the company's trends to the industry pattern may highlight that the company is serving a specialized market within the broader industry or that the company isn't keeping up well with its competitors. The forecast of such patterns may lead management to alter its strategies if such trends are unfavorable, or to concentrate more on a strategy that appears to be working well.
Forecasters may also analyze sales trends of individual products. This may include the use of price indexes. Such trends are important for understanding product life cycles and separating the performance of similar products (e.g., two different lines of shampoo from the same company) to evaluate strengths and weaknesses.
Forecasting involves more uncertainties than most other management activities. For instance, while management exerts a good deal of control over expenditures, it has little ability to direct the buying habits of its customers. Thus, even while sales trends depend on the vagaries of the marketplace, management must make a reasonable estimate of what the future holds in order to plan corporate affairs effectively.
The process managers or analysts go through to create a sales forecast is similar to this:
While forecasting is still neither effortless nor flawless, the gap between forecasts and reality has steadily narrowed over time. There are several ways that a company can improve the likelihood of creating an accurate sales forecast and using it effectively:
A variety of approaches can be used to surmise the future growth of sales. Some are highly dependent on statistics and mathematical relationships, while others are more inferential or speculative. The choice of approach depends on how accurate or precise the forecast needs to be, how long of a period it's for, the availability of past or supporting data, the funding available for forecasting activities, and other considerations.
In the causal approach, forecasters identify the underlying variables that have a causal influence on future sales. For instance, new computer sales generally have a direct influence on software application sales. In the consumer software market, other causal factors might be population growth, the expansion of computer-based activities such as electronic commerce, and trends in work and school practices like working at home and computer-based education initiatives. Still other more general factors might be the growth of personal income, employment levels, patterns in international trade (new market opportunities or new competitive threats), and so forth.
To first assess the market trends, the task of the forecaster is to establish (if it has not already been reliably by others) how these factors relate to one another and to sales of software. Some will have a direct (positive) relationship with sales, while others will have an inverse or negative relationship. In statistical parlance, the forecaster is identifying a set of correlations. Upon further examination, some of these correlations will appear causal (population growth causes higher sales), while others will be indirect or coincidental (inflation growth may cause both rising interest rates and rising sales, but rising interest rates may have nothing to do with rising sales). The sum of all this information is a formula or model that, given a certain set of conditions characterizing the underlying factors, will indicate the future behavior of sales.
On the next level the forecaster must assess the company's position in the industry and how that is likely to change over the forecast period. If no change at all were expected (i.e., it retains the exact same share of the market over the period) the company's sales would grow at the exact same rate as the broader market. Since this is uncommon, however, the forecaster must surmise whether the firm's recent or intended actions—as well as those of its competitors—will result in rising or falling market share. Again, there are a variety of causal variables to be considered, such as advertising expenditures, promotional efforts, new product introductions, and technological changes, to name a few.
Eventually, through data analysis, model construction, and statistical methods, the forecaster will arrive at a causal model of company sales based on external factors and internal actions. When changes in those factors occur (or are expected to), the implications for the company's sales can be determined by recalculating the forecast using the same model but different inputs. As this description suggests, causal approaches to forecasting tend to be complex analyses of a wide array of potential influences on sales.
A regression analysis is a specific forecasting tool that identifies a statistical relationship between sales, the dependent variable in the analysis, and one or more influencing factors, which are termed the independent variables. When just one independent variable is considered (say, population growth), it is called a linear regression, and the results can be shown as a line graph predicting future values of sales based on changes in the independent variable. When more than one independent variable is considered, it is called a multiple regression and can't be represented with a simple line graph. Regression analysis is related to correlation analysis, where the latter is concerned with the strength of relationships between the independent and dependent variables.
Another causal model is life-cycle analysis. Here product sales growth rates are forecast based on analysts' projections of the phases of product acceptance by various segments of the market—innovators, early adapters, early majority, late majority, and laggards. Typically, this method is used to forecast new product sales. Analysts' minimum data requirements are the annual sales of the product being considered or of a similar product. It is often necessary to do market surveys to establish the cause-and-effect relationships that signal the different phases of the product life cycle.
The most common noncausal approaches are time-series models, in which patterns are extrapolated from standardized historical data in order to reach a future projection. (Elaborate time series may also be used in causal models as well.) Analysts plot these patterns in order to project future sales. Because no attempt is made to identify and evaluate the underlying causes of sales patterns, the analyst implicitly assumes that the underlying causes will continue to influence future sales in roughly the same manner as in the past. Consequently, while it is easier to use and understand, this approach tends to be relatively simple and may not produce as reliable results as other methods.
One common method of forecasting based on time series is the use of moving averages. There are a number of specific methods that incorporate moving averages. All of these assume in some way that future sales will reflect an average of past performances rather than, for example, following a linear percentage increase trend. Moving average methods minimize the impact of random outcomes that could skew a forecast.
Exponential smoothing is a similar time-series technique. Rather than relying on equally weighed historical averages, however, exponential smoothing adds weight exponentially to the most recent values in the series. This assumes that the most recent figures are the best indicators of current trends and market forces, whereas older figures may represent an inaccurate or out-of-date picture of the sales trend.
Any sophisticated time-series technique also includes some provision for filtering out random noise or chance occurrences in the data that aren't part of the underlying sales trend. In a mathematical formula this takes the form of an error or noise term that is calculated into the forecast.
A number of approaches rely on the informed opinions of various individuals, who may consider past trends, causal factors, their personal observations, or any number of other factors to arrive at a forecast. Usually this involves asking a number of knowledgeable people from inside or outside the company what they expect will happen during the forecast period. The forecasters may be customers (intention-to-buy survey), sales staff, or outside industry experts who are familiar with the company and its competitors.
Aside from relatively informal internal surveys, perhaps the most widely known judgmental approach is the Delphi technique, which convenes a panel of (usually) outside experts who each come up with independent forecasts and then revise their projections until they reach a consensus position. Another important judgmental method is the program evaluation and review technique (PERT), in which optimistic, pessimistic, and most likely scenarios are developed (usually by one or more experts) and then weighed to produce an average expected scenario. A third general qualitative technique is called the probability assessment method (PAM), in which relevant internal staff members are asked to rate the probability of achieving a certain range or ranges of sales volume. The probabilities (given in percentages) are then translated into a cumulative probability curve that can be further analyzed to arrive at a forecast.
An intention-to-buy survey measures a target market's plans to buy a product within a given time period. Market analysts frequently conduct such surveys before introducing a new product or service. If it isn't a product they already purchase, respondents are given a neutral and reasonably detailed description of the product with the hope they will provide honest answers. When surveying the general public, care must be taken to ensure respondents don't provide unrealistically positive feedback on new product ideas, otherwise the results will be meaningless.
Such qualitative or judgmental methods are often preferred when (1) the variables influencing buying habits are changing or hard to determine, (2) enough data isn't available to support a statistical approach, (3) quantitative methods have given poor results in this forecasting situation, (4) the planning horizon is too far into the future for normal statistical methods to be useful, or (5) there is a need to consider technological breakthroughs which may only be in the early stages of development but will have impact during the forecasting period.
When forecasters first consider the broader market and then winnow it down to the company level, it is known as the indirect approach. When they only work with company data, it is called the direct approach. While indirect obviously lends itself more to causal analysis and direct more to noncausal, in theory direct and indirect approaches can be used in both causal and noncausal models. While for many sales forecasters the direct approach is most practical, it can be a revealing exercise to go through the indirect approach, since it requires that the forecaster consider the entire market potential for a product. Through this process the forecaster—or a recipient of the forecast—may discover unmet needs or other indications that the product's sales are performing well below potential.
Kress, George, and John Snyder. Forecasting and Market Analysis Techniques: A Practical Approach. Westport, CT: Quorum Books, 1994.
Mentzer, John T., and Carol C. Bienstock. Sales Forecasting Management. Thousand Oaks, CA: Sage Publications, 1997.