The text discusses the limitations of using the number of incidents as a measure for improvement in organizations. It argues that fewer incidents do not necessarily mean better reliability or safety, and more incidents can be beneficial if they lead to learning and improvement. Furthermore, setting targets for incident numbers can drive undesirable behavior such as underreporting incidents or arguing over severity ratings. The text suggests focusing on measuring tangible outcomes like service uptime instead of incident counts. It also recommends using incident data to improve processes and providing context when sharing incident metrics with leadership.