Three or so different ways, depending on camera, film and subject:
BW / LF: spotmeter shadows, check highlights and midtones to see what will happen if shadows are placed on zone III. Adjust exposure until feasible with home-mixed chemicals. Then forget my planned development and develop by inspection anyway. Or stand development in FX-2 if I have truly no idea of what it is.
Colour: Aim incident meter backwards over my shoulder, set speed and aperture accordingly (outdoors).
Old Folders: "Debatometer"! Look at the sky (or other light source), guess how strong the light is compared to "sunny 11" (empirically determined for 60° north latitude), open up for clouds, maybe close down for old uncoated lenses (less contrast due to flare), wiggle the dials a bit, look at the sky again, guess what the longest time I can hold the camera still is, set the aperture I want, look around for nice solid support. This sounds nowhere near as haphazard as it really is, but I get a surprising number of well-exposed negatives this way! I justify it as being "historically correct". Or should that be "hysterically incorrect"?