He is saying simple "spec numbers" are no good and he is right. You want to see full measurement graphs so that you can see the tested parameter varied. Of course this is what I do at ASR day in and day out. Here is the power measurements of an amp:
Note how I provide references of a mediocre and stated of the art amp at the same time to give you context. Without it, you can get lost in see of numbers he is showing you.
He then makes a bunch of other comments about power mattering at micro-watts which is neither here, nor there. First, he calls them "distortion" when in reality that part of the graph (sloping down) is completely dominated by noise. Distortion only takes over when the graph starts to move up which only happens near clipping (exception are amps without feedback or distortion-factory ones).
I also don't believe in this "first watt" business. Speakers have become smaller to fit in homes better over the years and this has made them much less sensitive. As a result, 1 watt is not going to do anything for you. For this reason, I have standardized on 5 watts as a low value power output:
Notice the comprehensive dashboard that not only gives you distortion, but also the amplifier gain and full spectrum of noise+distortion. The latter lets you apply perceptual analysis to determine audibility.
Bottom line, yes, run fast from any single value specs thrown out there without detailed data as I have presented. As a practical matter, for non-speaker products, you do need to stick to one source of measurements. In my case, I have tested over 1000 audio products of this class in the last few years. So you can compare with confidence.
For speakers (and headphones to some extent) we have quality measurement standard in the form of ANSI/CEA-2034. With this data, you can indeed compare one source to another. This is what it looks like:
In on graph you get on-axis and (important) off-axis response. And directivity.
You don't get distortion so I have added that with two (and sometimes three) standardized levels:
I have over 200 speakers measurements (I think) and others have posted measurements using similar scheme. That hole then is being filled nicely although the process remains more expensive than for electronics.