An exercise that was useful for me was taking 2 batches of 10 cases and experimenting with different brass prepping practices. Use the best load you currently have and only vary the case prep. Start with the same for both sets to see how well they repeat. That's a first cut at the resolution of your test. As your loads improve, repeating the 10x10 race may give further improvements now that the next layer is out of the shadow of the larger problems.
Eventually, the problems become small enough that you'll need to make or hunt for examples at the extremes and race them against each other. Heaviest vs lightest cases, runout up against runout down.....
