The Kill-a-Watt is very good. I even tested one against a Fluke wattmeter and it agreed spot on with a 1000W heater. It's at low power where it starts losing accuracy.
Here's an interesting wattmeter circuit for building one into a device:
Wattmeter_from_Pease_after_Nelson
Probably not terribly useful on a refrigerator (very small changes in instantaneous power use, just differences in duty cycle) but might be quite useful on devices that vary greatly in power usage.