When General Motors was facing increasing competitive pressures, higher quality demands and a sluggish economy, the giant automaker was unable to effectively analyze the productivity and throughput of its manufacturing operations. Many of its factories during the late 1980s and early 1990s were missing production goals, working unscheduled overtime and experiencing high scrap costs. Although GM overall had excess production capacity at the time, production bottlenecks and other problems were causing the company to lose money even on high-demand products. In addition to creating problems for existing products, the lack of an effective throughput analysis capability was impeding the launch of new products, resulting in lost sales. In 1991, GM reported a $4.5 billion loss – a business record at that time.
Recognizing the broad scope of the problem, GM responded with a three-pronged analytics-based effort focusing on production line data collection, throughput modeling and algorithms, and throughput improvement processes. In tackling the production data collection component, GM recognized the importance of limiting data inputs from the massive universe of production data available, to avoid bogging down the analytical processes. The goal was to develop models and algorithms with modest data requirements that still produce meaningful results. Repeated trial and validation efforts established the appropriate data inputs in the areas of workstation speed changes, scrap counts and