Foresight: The International Journal of Applied Forecasting, Issue 38, Summer 2015, Pages 5 – 8
Some forecasters might envy those who make demand forecasts for products that have been around for several years. After all, they have access to plenty of past history and data that follow well-rehearsed trends and seasonal patterns. They also have frequent feedback on accuracy and established software products for support. Some forecasters even reap the benefits of information sharing with suppliers and customers. However, two factors can turn accurate demand forecasts into a tough proposition: intermittent demand and promotions. Intermittent demand occurs when some periods have zero demand; this poses a challenge because commonly used forecasting methods, like single exponential smoothing, can lead to seriously biased forecasts when demand is intermittent. Promotions cause problems because they disrupt regular demand patterns. Forecasting methods that aim to detect regular patterns often cannot cope with the resulting disruptions and have no way of predicting the sales uplift that a promotion might achieve.