Using No Data Alerts
It's possible to set any datapoint to create an alert if no data is reported or able to be collected. The most common uses of this feature is to allow an alert to be raised to demonstrate the following:
A collection method has failed
This is generally put on a single instance datasource on a datapoint that should always be responsive in order to let the user know that an entire collection method is not working and should be checked. For example, WinSystemUptime can be set to create a warning alert if no WMI data is being returned for the SystemUpTime datapoint.
An important instance has disappeared
It is generally not recommended perform this on multi-instance datasources because a number of No Data alerts would be generated from one fault.For multi-instance datasources where there's the potential for important instances to disappear, one may consider setting a no-data threshold. This would most likely be done when the instances are entirely distinct from one another. (perhaps port-discovered instances where the user was running more than one DBMS or web server)
In either of the above cases, the datapoint chosen is generally one without a threshold. This allows for customization of the alert message and for message body to reflect the lack of data, and not a measurement exceeding a threshold (you wouldn't use the ##VALUE## or ##THRESHOLD## tokens in these).