In stats methodology research, do both simulations and real data analysis for the simplest version of method, then iterate.

Don’t do increasingly complex methods in simulation, and then, at the very end, try it on the real data. It probably won’t work, and, even if it does, it probably won’t have much impact.