This paper analyzes the basic performance of 27 automobile engine lines operated by 18 companies on three continents, based on questionnaire data gathered in the Spring and Fall of 1995. Engine plants differ from assembly plants in being very capital-intensive. Thus a traditional ?hours/engine? metric of performance is inappropriate. Here a composite cost comprising labor and amortization of capital, accounting for downtime, is used to compare plant performance. We find that performance varies widely, even for similar engines. Cost drivers comprise number of workers, capital invested, and efficiency (fraction of scheduled time actually used for production). The drivers are in turn driven by external factors out of the plant?s control and internal factors that are under its control to some degree. We find that about half the variance in cost is due to the external factors, such as number of cylinders, utilization of scheduled time, and number of variants of engine made (the last loosely related to age of the engine family). Internal factors such as work in process inventory (strongly) and age of the workers (somewhat) drive cost. Downtime, the reverse of efficiency, is itself divided into scheduled and unscheduled downtime; the former is driven largely by number of variants while the latter is driven to some degree by the age of the family. The results of this study include a methodology to estimate the cost of variety. Statistical analyses are used to calculate the additional cost of machining blocks ($4.92 more per block, $15 million extra investment, 9 additional workers and -4?40 operating efficiency associated with one additional square root of number of variants). This methodology can be extended to create a cost of variety for an entire engine.