There's not really an exact science to it and it depends on the design specifics for your CPU and GPU coolers. The best way I've found is to just attach it to the back of the heatsink contact plate if you can get to it - if not, then to the closest cooling fin.
Once you've done that, monitor the difference in temps reported by the probe and by the component (using HWMonitor or other software that pulls temp data from the mobo sensors). It'll vary depending on the load, so you'll want to graph multiple load/temp points. Then factor that differential into your fan controller profiles. For example, if you want your fans to run at 50% at idle, if your CPU idles at 30C according to HWMonitor and your probe is reporting 25C, set your...