Start with an observed event

Keep the observation separate from a proposed cause. Ask what evidence would distinguish the competing explanations.

Choose a change you can review

Assign an owner and a review date. Keep the action small enough to establish whether it helped.