Noisy measurement

This paper tackles the challenge of ranking noisy measurement data sets while accounting for ties, which can lead to non-transitive relationships and various valid ranking outcomes. It introduces three methodologies for partial ranking. 

The first generates a partial ranking with a flexible number of ranks, while the second refines this by reducing the number of ranks, addressing ties caused by overlapping and well-separated variants.

 The third methodology further optimizes the ranking, producing the fewest ranks possible. These approaches show potential
in identifying performance differences among business process variants, demonstrated through the Purchase-to-Pay (P2P) process. This analysis provides a foundation for developing models that automatically differentiate process variants, even when performance metrics are unavailable.