The focus of this empirical investigation is twofold. First, an examination of formal hierarchical forecasting system structural parameters is reported. Four formal experimental factors used to define the hierarchical forecasting system are examined. These factors are evaluated in a controlled comparison by means of parametric statistical analysis. Second, a comparative analysis of formal hierarchical forecasting system performance with (1) an informal hierarchical strategy, and (2) a traditional forecasting approach is presented. The best performing combination of factor level settings identified in the examination of formal hierarchical forecast strategies is subsequently evaluated in the second part of this analysis by means of paired difference comparisons with (1) a randomized, informal approach, and (2) a common forecast strategy over the same time series data. Traditional measures of forecast accuracy are used as dependent performance criteria. 0272-6963/95/$09.50

Constrained forecasting has become a popular approach to the complex task of predicting future sales. Mixed performance results have been obtained by companies employing constrained forecasting systems software packages which probably reflects the lack of any clear system implementation guidelines. A number of important operational parameters such as the number of product families, criterion used to define product families, number of system levels, etc. must be specified. This study reports on an investigation of two of these parameters: product group size and appropriate grouping criterion. Different tests are performed with the constrained forecasting system and compared to a traditional forecasting approach using product data from the