The declarative style of a Makefile (or substitute your own language-of-choice’s implementation of *akefiles) lends itself well to scientific processes, where a reproducible method is crucial. I recently found this out when analysing some data, which consisted of the following:

1. compile the source code of the programs I was using
2. use these programs on the same data
3. plot some summary plots
4. show the plots

With the declarative style of Makefiles I could rely on the latest results no matter what changed in the analysis path, for example the plot scripts used in step 3. were listed as a dependency of the step run to make the plots so when the script changed, the plots would be recreated.