More changes do not create more information
Changing several components or setup details at once may produce a different result, but it does not explain why. I keep a baseline, change one planned variable, and record everything else that could affect the comparison.
Define the question before the test
A useful question has a measurable answer: did consistency improve, did the point of impact repeat, or did the process become easier to control? Decide that before seeing the target.
Keep safety outside the experiment
Published data, component-manufacturer guidance, and equipment limits are constraints, not variables to explore. If anything looks abnormal, stop and verify it. No score or graph is worth pushing past uncertainty.