01
Frame the symptom
Choose one interaction and describe the visible failure in frame-time terms rather than as a vague feeling of slowness.
02
Trace before tuning
Capture CPU work, allocations, traversal, drawing, and composition around the same reproducible gesture.
03
Verify the trade-off
Repeat the trace after each change and record what moved, what did not, and which cost was exchanged for another.
Working rule
Make the next measurement cheaper than the next argument.
Keep a minimal reproduction close to the observation. Change one variable at a time, preserve the failing path, and prefer before-and-after evidence over a broad rewrite.