It is very hard to measure responsiveness (which is basically the only thing zen claims to be better at) without a lot of time investment and a high speed camera for frame analysis.

I guess a possible setup would be to run a multicore synthetic workload that takes most of the system resources and measuring the start time desktop applications.

But nobody has done it in a “scientific” way as far as I can see. Besides some hobbyist anecdotal observations:

https://reddit.rtrace.io/r/linux/comments/fh0hc2/its_time_for_desktop_distributions_to_adopt_a/

https://forum.garudalinux.org/t/kernel-benchmarking-results-zen-cacule-tkg-bore/18831/8