The CPU utilization metric is frequently utilized as a trigger for scaling actions because it effectively represents the load on compute resources. If CPU usage exceeds a certain threshold, it indicates that the application may require additional instances to handle the load, leading to a scaling action. Network throughput is less commonly used for horizontal scaling as it does not directly indicate the load on the compute resources. Lead time and memory utilization might also influence scaling decisions, but they usually play a secondary role compared to CPU utilization in triggering horizontal scaling.