We report results from a multi-year empirical study on the impact of product variety on automobile assembly plant productivity. Three sources of data were used to examine the research questions and triangulate the findings. First, we conducted field research in more than 20 assembly plants worldwide to study how they cope with product variety through manufacturing flexibility. Second, using time series and cross-sectional data from GM's Wilmington, Delaware plant, we examined the effects of

more »

... uct variety and the variability in work content per car on various performance measures. Consistent with the field evidence, the time series analyses indicated that greater variability in option content has a statistically significant adverse impact on total labor hours per car produced, overhead hours per car produced, downtime in the assembly line, minor repair and major rework, and inventory as a percentage of cost of sales. We also found that work stations with higher variability in option content have significantly higher slack resources to protect against variety-related defects and downtime, introducing an additional fixed cost of product variety. Finally, we used simulation to measure the impact on assembly line downtime of variability in work content per car. The simulation study confirmed the results of our statistical analysis. In the last few years, auto manufacturers worldwide have been retrenching on the level of variety they offer, in part to reduce costs during the recent recession. But whether variety has been increasing or decreasing, what remains constant is a general lack of understanding of variety's impact on manufacturing costs. The accounting and operations management literatures have emphasized the potential costs from increasing variety (e.g., Skinner, 1974; Hayes and Wheelwright, 1984; Miller and Vollmann, 1985; Cooper and Kaplan, 1991) . Researchers in these fields argue that higher product variety creates considerable challenges for manufacturing operations. With an increasingly complex product mix comes additional parts, less accurate demand forecasts, greater inventory and material handling, additional setups, more complex scheduling and task assignment, greater chance of quality problems, and increased supervisory requirements. As a result, greater resources must be committed to handle the increasing number of transactions and manufacturing contingencies and ensure the smooth operation of the plant. Although greater product variety is widely believed to be associated with higher manufacturing costs, empirical studies show mixed results. Research using the Profit Impact of Marketing Strategies (PIMS) data base found that significant market share benefits accrued from broader product lines, but observed no relationship between self-reported production costs and product variety (Kekre and Srivasan, 1990). Cross-sectional accounting studies by Foster and Gupta (1990) and Banker et al. (1993) also found little association between direct measures of product mix complexity (e.g., number of parts, number of suppliers, breadth of product line) and manufacturing overhead costs, after controlling for direct labor costs. In contrast, a time-series analysis of textile plant production costs by Anderson (1995) found that more heterogenous product mixes increased manufacturing expenses. In the automobile industry, MacDuffie, Sethuraman, and Fisher's (1996) cross-sectional examination of assembly plant productivity indicated that differences in parts complexity (e.g., number of suppliers, number of parts) negatively affected productivity, while model mix and product options had little impact. However, in plants employing "lean" production practices, parts complexity was not significantly correlated