Jonathan Thatcher, CSCP | January/February 2012 | 22 | 1
Extrinsic forecast models take a wide perspective
Reader H.Y. writes, “My team’s forecasting model is not very sensitive to socioeconomic trends. How can I use these data in forecasting?”
To achieve your goals, your team should engage in extrinsic forecasting. “Extrinsic” means coming from the outside. At most organizations, this refers to the wider economy. Extrinsic forecasting is most useful for evaluating the external conditions common to supply chains, large markets, and geographies. The forecast horizon and scope are necessarily long and broad, as the model presents a bigger picture than even traditional long-range forecasts—one that reveals the enablers and inhibitors to overall demand and supply.
Extrinsic forecasting depends on the aggregate picture. Even in areas where it is difficult to accurately forecast individual items, it is reasonably possible to forecast the aggregate. Aggregate trends tend to move in repeating, long-running cycles.
Common elements in extrinsic forecasting models include macroeconomics factors, such as inflation, Examining these factors can help anticipate changes and find the optimal timing to introduce products and modify prices. Extrinsic models also include macro-social demographic data, such as average ages, generational differences, preference patterns by region, and trends in immigration and emigration. These factors can influence broad customer decision-making patterns.
In industries heavily dependent on information technology, macro-technological change and development is an extrinsic forecast component. Technical advances can change fundamental processes, standards, and costs, while innovations from labs and development pipelines can suggest when new releases will hit the marketplace. These factors affect product development schedules, release cycles, and even remanufacturing demand for old components.
Begin by collecting macro-level data of interest, such as economic or demographic trends. Often, this information is freely available on national economic, trade, and statistics departments. Other factors to examine include import and export trends, sales per employee, and total production value. These serve as high-level baselines to compare your organization to historical averages.
With these data in hand, consider using the Delphi technique as the first step in your forecasting. Also known as “panel consensus,” this consists of polling experts representing a cross section of locations, responsibilities, and cultures about subjects such as trends in geographic supply and demand. The experts independently submit their answers and reasoning to a panel manager, who in turn asks the panelists to evaluate the likelihood of each response. Panel members may begin to see new perspectives, and the process can cycle a few times before reaching a shared conclusion that a response has a high probability of occurrence.
While the aggregate view is not ideal for short-term or precision forecasting, even short-lead-time products and services can be sensitive to extrinsic factors. For example, weather is a common extrinsic forecast element. If the big picture is the only concern, then forecasting would only examine the four seasons as repeating weather cycles. But, on any given day, an unexpected rainstorm or heat wave can shift customer demand up or down.
The same holds for economic factors: For example, even during periods of economic slowdown, there may be areas experiencing expansion.
Blending extrinsic forecasts with local or short-term forecasts helps create a more accurate view of the near term or the local picture. Try to gather meaningful, representative samples of both quantitative and qualitative data to create useful extrinsic forecasts. Compare the findings. Look for correlations, historical patterns, and useful data points that can refine projections.
Further refine your forecast by considering questions such as: What story do the extrinsic factors tell? Where has the organization arrived from? Where is it now? Where might it end up if long-running trends stay in place? How likely are those trends to remain in place? Review your analysis regularly, and continue to update it every year, quarter, and month. Eventually, you will see the past, present, and future of your forecasts as a single, unified concept.
Jonathan Thatcher, CSCP, is director of research for the APICS professional development division. He may be contacted at firstname.lastname@example.org.