If you are looking for 1uV of resolution then yes you can use averaging, however to obtain 1 uV your are going to have to average over 5000 samples. Remember, doing so will reduce your overall measurement bandwidth.
If you require a large dynamic range you may be better off using multiple channels to monitor the same signal. The first channel would be driven with a buffer (or small gain depending on signal amplitude) and would monitor the input when the signal is large. The second channel would be configured with a gain say of 32 or 64 (30 or 36dB) and would monitor the same signal when the input is below 1/32 or 1/64th of the input range. You can then use oversampling/averaging of both outputs to reduce any additional noise introduced into your measurement by the noise gain of the input signal conditioning.