Issue link: https://industrialmanagement.epubxp.com/i/285723
30 Industrial Management on only one forecasting method. Different methods must be used and tested for their validity. Both statistical and judgmental models of forecasting should be integrated to measure their effectiveness and accuracy on a regular basis. • Know your market: Knowledge of market and industry is crucial and can assist in identifying demand patterns and seasonality. • Integrating business intelligence with industry information: Business intelligence can be a useful source of market/industry information. This will help in making better strategic decisions. It may also provide early warning signals of the competitors' initiatives, technological develop- ments, regulations and customer needs. The organization will be able to learn from the success and failures of other companies in the industry. • Establish meetings: There should be regular meetings with those involved in the forecasting process. This will help to determine the needs of the forecast users. • Develop management trust and understanding: Management trust and understanding can be developed by involving management in the forecasting process. This will help to understand the internal forces that can affect the assumptions of the model. • Continuous training: The success of any forecasting model depends a lot on whether a training process is in place within an enterprise. Training helps in various ways, such as developing a balance of information power and providing a clear under- standing of assumptions and limita- tions associated with forecasting techniques. A journey in forecasting efficiency Bayer HealthCare is a pharmaceutical and medical product giant. Despite this, prior to 2001, many products at Bayer HealthCare Consumer Care EU had a built-in bias of forecasts that wound up overesti- mating or underesti- mating sales. The question was how to mitigate these biases and generate more accurate forecasts. To determine the forecast perfor- mance, three different error metrics were used: absolute percentage forecast accuracy; percentage forecast error; and standard deviation of 1 percent forecast error. The team found that forecasts for certain products included some inten- tional bias that resulted from Bayer HealthCare's organizational framework and incentive plans. Some products were predicted to sell too much because of budgetary bias. Adjusting a forecast to a realistic level would have meant that the forecaster had acknowledged that the budget would not be reached. Some products were predicted to sell too few units because of the incentive plans. These incentive plans influenced sales workers to lower forecasts in order to maximize the payout of bonuses. With such observations, Bayer HealthCare Consumer Care EU decided to restructure the forecasting process. It laid out new practices and also decided to use the CPFR approach. The results were positive: an improvement in overall forecast accuracy. Overes- timates of sales forecasts declined by as much as 5 percent, reported Heiko Petersen in "Integrating the Forecasting Process with the Supply Chain: Bayer HealthCare's Journey," from The Journal of Business Forecasting Methods & Systems. In addition, the forecasting bias was reduced from 15 percent to a remarkable zero percent. Obviously, a forecasting process is vital for a company's bid to reduce costs, increase profits and satisfy customer needs. Forecasting, as we see in the Bayer HealthCare example, is all about managing the truth and not justifying personal objectives and targets. Hence, it is a process that should be distinguished from the decision-making process. By estab- lishing a robust forecasting process and by practicing proactive control, forecast practitioners can avoid pitfalls. It will be helpful to organizations to set up workshops to discuss issues related to forecasting biases and their consequences. In light of all potential biases, managers should put in place a forecasting process that offers trans- parency of both the forecast and its inputs while providing a robust set of benchmarks and other data to use in challenging the forecast. In addition, a good forecasting process will accept that errors will exist; so particular attention should be directed toward developing appro- priate plans to manage errors. Failure to do so will result in disappointment and encourage company personnel to override the forecasting process, adding in more bias that assuredly will trigger even higher forecast errors. v Particular attention should be directed toward developing appropriate plans to manage errors. IM MarApr 2014.indd 30 3/24/14 11:02 AM