I think that time-wave data is helpful in trying to visualize the motion of the equipment, especially if the amplitude is in terms of displacement.
If you have an unknown wave form where you can't easiy pickout individual motions or waves, that's where the spectral data helps. Look at the specta, that shows you the frequency peaks, and from there you can go back and try to pick out the wave forms.
Some hardware/software systems let you go back and filter out certain frequencies and review the wave data with some of the "nosie" removed from the wave form.
As with any vibration analysis, you need first to understand the nature of the equipment, and what it's normal vibration signature is, and how abnormalties show themselves (like misalignment or oil whirl) and that way any deviation in the waveform data or spectal data can be detected...and data trending is the way to do that, which wave form comparison is not really practical...it's best to compare historical spectral data.