Benchmarking semiconductor manufacturing performance using a pairwise-comparison method

Benchmarking performance against competitors is an increasingly popular activity within the semiconductor industry. Owing to the high variability in manufacturing metrics, a simple monthly metric score is usually insufficient to accurately benchmark performance. However, in the rapidly changing semiconductor world, data which are one to two years old may be irrelevant to the current situation. We propose a simple ranking method based on a method of paired-comparisons for unbalanced data which gives a good ranking and is also reliable under fluctuations in the data. Using semiconductor manufacturing performance data from the Competitive Semiconductor Manufacturing Study at the University of California at Berkeley, we give an example of this method and compare its reliability and accuracy against other techniques through simulations.