|
|
|
|
|
Comments or Suggestions?
|
|
|
|
|
|
|
By Mike Rose, Director of Development, SalesLobby.com,
Sales Compensation Consultant, The Alexander Group, Inc.
Forecasting in challenging economic times is critical to success. Here are some concepts and techniques to ensure accurate forecasting.
Forecasting has always been somewhat of a black art and in good economic times as long as forecasts were optimistic that was enough. But in challenging economic times, it becomes increasingly important to focus on accurate forecasts.
This is true not only because it ensures a common understanding of sales goals throughout the organization but also because it is important to communicate realistic expectations to the investment community. In pursuing this goal, the sales organization plays a critical role in providing information on future revenue. Here we discuss a few approaches and techniques that provide effective means of achieving accurate forecasts.
Revenue Analysis
What are we typically forecasting? Total Revenue. To accurately understand an aggregate quantity like total revenue or sales, it is typically helpful to break it down into its constituent parts. By understanding the different sources of revenue, we can build models that incorporate an understanding of the behavior of these sources. These models may predict revenue by source that can then be rolled-up for a global picture.
The basic source of revenue is customers and the means of obtaining revenue is delivering product, hence, an understanding of customers and products is critical. As a result, we generally build models by product that incorporate an understanding of the product market including customers needs and buying habits. This generally is the kind of research that marketing provides at the market level and sales provides on specific customer and prospect buying expectations.
Figure 1. The Product Life Cycle
Products, like organizations and living organisms, follow a predictable pattern throughout their lives. Because a business is comprised of products, accurate forecasting requires an understanding of the location of the product in the life cycle. In the beginning, they experience a fast growth curve that begins to level of as they reach maturity. Flat or modest growth characterizes the mature period followed by a gradual decline toward obsolescence.
Hence, we observe three basic periods: growth, maturity and decline. While the shape of the life cycle is predictable, the actual length of each phase and height of the curve depend on many factors. In the following, we focus on analyzing product sales using to general methods:
• Heuristic Models – Intuitive and Frequently Simpler to Construct
• Statistical Models – Analytically Rigorous and Complex
Heuristic Models
Heuristic models rely on understanding the basic business cycles that drive business sales. These cycles include the life cycle, seasonality, organizational influences including reporting as well as activities in the market place. By identifying and isolating each of the business cycles, we decompose total sales and then using the period and amplitude of these cycles to extrapolate sales numbers typically by month. We can create a fairly accurate forecast of the period of interest simply by repeating these cycles and adding them back together.
For companies with many products in different phases, it helps to use information about the typical growth cycle and other business cycles to predict the behavior of new products.
The strength of the heuristic model is its simplicity, it can be built in Excel, and reliance on well understood business dynamics. Even if we have no data for product sales, we can still use basic product and market cycles to make predictions. Its weakness is the lack of statistical measures of error and model strength.
In addition, the heuristic model requires an understanding of the environment and influencing factors sufficient to be able to hypothesize (and test) theories about each identifiable factor (or cycle) in de-constructing the historical data. So another key benefit of heuristic analysis is the ability to account for changes in the factors that influence the underlying cycles. Such knowledge is important in creating accurate constructions of forecast data.
Figure 2. Business Cycles
Statistical Models
The most commonly used statistical models consist of the venerable regression analysis. In this approach, one or more quantities may predict values of another quantity. Because forecasting is basically time series analysis, that is, it relies on using past values of a single value to predict future values, more advanced methods are generally used including auto-regression.
An auto-regressive model is simply a regression model whose predictive variable (product sales) is regressed against past values for the same variable. Hence, auto-regressive models are ideal for time-series analysis (where we have data by time period for previous sales). This approach produces models whose output variable is a function of former values of the same variable. In addition, these models require software like SPSS or SAS. Typically, we explore a few models providing statistically significant coefficients and high R-square (model strength). We then apply these models to form the basis of our forecast.
The strength of the AR models is their statistical robustness and application to somewhat random information (i.e. the stock market). Their weakness is their reliance on historical data (typically at least a year or more) and resistance to intuitive interpretation (i.e. What is the model saying in business terms?). The statistical model looks only at the historical data points (and so assumes that all influencing factors are reflected inherently in the data itself). This isolates the analysis (and analyst) from any requirement to understand the influencing environment and factors. It could therefore be argued that this is the more simplistic approach. The statistical approach assumes the influences inherent in the historical data will remain constant and so may have more difficulty in accurately predicting future data points in evolving or shifting operating environments.
It is not uncommon to use both the heuristic and auto-regressive approaches to come up with a few different forecasts just to get a feel for the variability and range of possible predictions. When all models agree closely, we can be reasonably confident in our results. If we see significant disagreement amongst our models, we had better do some more homework.
Marketing and Sales
No forecasting model should be considered complete without the input of the marketing and sales organizations. Marketing can provide critical information around the business environment including competitive entry, governmental regulations, etc. Incorporating information based on our customers as well as products is best left to the sales organization. In these challenging times, leaving out direct customer information from the sales organization is a prescription for unrealistic expectations.
Each sales person can provide forecasts by product and customer based on business experience. In fact, many simple tools exist today to facilitate this process (including Sales Forecaster). This data can then be compared with and incorporated into the analytical forecasting approaches described in the previous sections to ensure that our theoretical techniques mesh with the real world.
Reference
Wheelwright, S.C. and S. Makridakis.1985. Forecasting Methods for Management, John Wiley & Sons, New York.
|
|
|
|
