You need to be familiar with the angle of view of the meter, and to be able to tell what the meter is actually seeing.
You can then look at that portion of the scene, and evaluate for yourself how much the average reflectance of that portion deviates from a middle/18% grey.
In essence, I look at the scene and say to myself something like: "that looks to average out at about two stops/zones brighter than a middle grey. So I better take a reading and then open up two stops."
I prefer to use incident metering.