Prove that staging actually predicts production — same config keys, same migrations, same feature flags, same integrations — and catch the differences that make a green staging run meaningless. Use before trusting a staging result, after a deploy that behaved differently than expected, or when something worked in staging and broke in production.
