Legacy software is a prominent bottleneck in modern industry: it is hard and costly to maintain, yet contains valuable knowledge not available elsewhere. Since manual rejuvenation of such systems is costly and cumbersome, meta-programming techniques can be used to decrease this effort. Generic transformations, however, do not take domain knowledge into account and are therefore not satisfactory. In this talk, I will discuss an ongoing project I’m conducting together with Philips Healthcare, in which we perform a significant refactoring on a part of their codebase. I will highlight how we have broken up this refactoring in smaller transformations, and how we evaluate the quality of each intermediate step.