One thing worth trying is a side-by-side comparison around personal. Keep the input the same and change only the part you're testing. It makes the difference much easier to spot. That usually gives a much cleaner signal than changing several things together.
