 |
|||
|
GfK Healthcare | August 2010 | |
|
|||
Good Practices in Forecasting and Good Forecasting in Practice – Part One  By Rudiger Papsch, Managing Director, GfK HealthCare Asia The ability to make accurate forecasts is highly mission critical for any organization. This probably applies to pharmaceutical and medical companies even more than other companies, as in the health care industries the development cycles are longer and the development costs are higher. Forecasting comes into play whenever things need to be planned ahead of time. In today’s professionally managed organizations it is hard to think of anything that would not require careful planning, and for this reason forecasting is important for virtually all functional areas of health care companies, including strategy, research and development, marketing, production and finance. However, inaccurate forecasts result in either lost opportunities or false investments, which both can lead to substantial financial losses for a company. Numerous examples of poor forecasts can be found in the literature and are passed on within the health care industry. A promising new product was never launched because the market potential was underestimated; expensive raw materials and production capacities were bought because the sales during the winter months were overestimated; and a suboptimal marketing strategy was chosen because the counteraction of the key competitor was incorrectly predicted. To be fair, there are many challenges and complexities in forecasting. These can be grouped into two categories: technical challenges and organizational challenges. Technical challenges refer to the details of making the forecast as such: Which forecasting methods are chosen; on which external data sources is the forecast built; and how is the forecast evaluated? If those challenges are not answered appropriately, they will usually lead to inaccurate forecasts. Organizational challenges refer to the way the forecasting process is implemented in the organization and how members of the organization deal with the forecast: Which stakeholders are involved in the forecasting process and how are they involved? How is the forecast communicated within the organization? Those challenges also can have a negative impact on the accuracy of the forecast, but even more they can make the forecaster’s life difficult. How can these challenges be overcome? Substantial advances in forecasting approaches and techniques have been made in the past decades that address these challenges effectively. However, most of these improvements are hidden inside the academic literature on forecasting and seldom find their way into practice. This article summarizes these advancements in forecasting into five simple guidelines that will support you on tackling the technical and challenges of forecasting. In the September issue of Pipeline, we will focus on the organizational issues and how to put good forecasting into practice. Good practices in forecasting – how to tackle the technical challenges 1. Decompose the forecasting task: Typically forecasting is a complex matter because each forecast involves many factors and variables. Furthermore, better information is available about each of these single elements than about the topic as a whole. Therefore an effective yet simple way to increase the accuracy of your forecasts is to divide the forecast into smaller units which are solved separately and are then combined to give the overall forecast. For example, to estimate the total number of cars in the United States is a fairly challenging task. If this estimation is structured into population size, members per household and cars per household it becomes a manageable task that will result in a much more accurate estimation. 2. Select a forecasting method that fits the forecast: The "best" forecasting method depends on the situation. A method that might be a good choice for one specific forecast might be a poor choice for another. In practice this advice is rarely followed by forecasters; they tend to use a single method for all kinds of forecasting problems. If you happen to have approached all forecasts so far with a single method, you may want to enrich your repertoire with additional methods that allow you to select the best method depending on the situation. 3. Prefer simple quantitative techniques over more complex ones: Evidence from various studies on time-series, econometric models and diffusion models show that simpler quantitative methods are usually as accurate as complex ones. Surprisingly, often simple quantitative methods are even more accurate. This is the case because more complex methods have more variables and therefore allow more possibilities for poor data and assumptions to enter the forecasting model, while at the same time leading to only marginal accuracy gains through their higher sophistication. In addition, simple methods have a better buy in from the stakeholders. 4. Be aware of the strengths and limitations of qualitative forecasting techniques: Judgmental forecasting methods have their firm place in forecasting and are useful for many forecasting problems. However, studies demonstrate that they are less accurate compared with quantitative methods if both methods can be used for a particular forecasting problem. The main reasons for this limitation are the bias of the forecaster, natural cognitive limitations of human information processing and, quite frequently, overconfidence on the part of the forecaster in being able to give an accurate "estimation" of the situation. It is therefore advisable to use judgmental forecasting methods only if you are lacking sufficient objective data. In general, the more structured judgmental forecasting techniques (e.g., Delphi method) should be preferred over simple, straight-forward estimations of experts. 5. Evaluate your forecasts: The complexities involved in forecasting also mean that there are many opportunities for you to improve your forecasts further. Comparing your forecasts with the actual outcomes and understanding the reasons behind deviations is the most effective way to improve them. Keeping a track record will also help you build your reputation as forecaster. Various measures for forecasting errors are available; the Mean Absolute Percentage Error (MAPE) is often a good choice. Want to learn more on this topic? Please contact:  |