Product forecasts are critical input into procurement, inventory, marketing decisions etc. The use of human judgement is common in the real-world forecasting practice. Human intervention occurs mainly to incorporate contextual information. The literature suggests that a forecasting support system (FSS) that systematically guides the forecaster in applying judgement can improve forecast accuracy. Guidance is the core component of such an FSS. A behaviourally-informed FSS (BIFSS), as defined in this thesis, is an FSS that aims to provide systematic guidance to inform the judgement of a forecaster.
This thesis firstly investigates the impact of promotions on product sales and judgemental forecasts using industry data. Then, a novel conceptual framework for developing a BIFSS is presented based on the literature of decision support systems (DSS) and judgemental forecasting literature. Decisional guidance as a crucial element in this framework is selected for further investigation. A lab experiment is employed to examine the effectiveness of two types of guidance, interval guidance and adaptive guidance. The moderating impact of promotions as a major contributor to the complexity of forecasting (shown by using the industry observations) is also considered in the experiment design. Task complexity also varies by changing the noise level in the time series.
The results confirm the effectiveness of guidance types in improving forecast accuracy. However, there was not a significant difference between the guidance types. It was also found that providing multiple guidance types does not necessarily result in more accurate forecasts. My analyses provide evidence that guidance is particularly effective under the most complex task setting. The positive effect of guidance under less complex settings is not significant. Providing multiple guidance types can better help forecasters overcome a higher complexity than only providing one guidance type.