Often, a metric is useless in isolation and needs to be evaluation in combination with another. For example, a high Signal Waits Percent is just noise without taking into account the Signal Wait Time (raw).
How much should I care about high Host CPU Usage if Instance CPU Utilization and O/S CPU Utilization (%) are normal? If the host has many non-SQL guests, I probably don't care if those other guests are getting shorted on CPU as long as my guest is getting what it needs.
Being able to base an alert on multiple metrics would help cut down on "noise" alerts and make more meaningful situations more visible.
