When production breaks, several engineers debugging in private messages without a plan makes things worse. Incident response gives clear roles, like an incident commander, and a fixed order of work: assess, mitigate, resolve, then review. Use it for anything that hurts users, such as a bad deploy that breaks checkout, where the first move is to roll back. Do not chase the root cause while users are still down; stop the impact first and communicate early.