This is a very broad subject.
What matters in a receiver is how well the signal is actually recovered, and that is directly impacted by the signal to noise ratio.
There are many sources of noise and fading and the first one (that we cannot get rid of) is thermal or Johnson-Nyquist noise.
This is defined as \$ P = kTB\$ where P is the power in Watts, k = Boltzmann's constant (\$1.38 \cdot 10^{-23}\$) T is the thermodynamic temperature in Kelvin and B the bandwidth of the signal in Hertz.
Then the sensitivity of a receiver (without using special techniques that I will touch on) is at a power level of thermal noise plus the noise figure of the receiver circuitry.
The higher the signal to noise ratio, the lower the bit error rate (for digital signalling) or the less noisy the analogue signal is (for an analogue domain). Note that although a signal may appear to be crystal clear, some noise till exists; you just cannot hear it. This noise level applies for any specific given type of modulation.
Different forms of modulation have different channel capacities vs Signal to noise ratio.
There are many types of signal encoding.
There are methods of recovering signals that are below the noise floor by using averaging techniques; if a signal with a relatively narrow band is continuously present, and the frequency range of interest is accumulated over time, it will rise above the noise (which is broadband and flat).
GPS is one such example of this.
Hopefully, this very broad brush look shows why there is significant knowledge required to determine the best metric for a given signal link.
