This is an interesting article! As a market researcher, I could really relate to this quote:
“It happens all the time: The Company has a problem. The Company commissions research. The research comes back, and tells The Company something unexpected, or unwelcome. The Company ignores the research.”
I’m still a little vague on why we need normalization. Isn’t that what the volume control on our systems is for?
I sometimes feel that I’m in a house of mirrors. What am I hearing? There are so many layers to the onion–and I can’t figure out which layer is doing what. Oh well 
Many who use playlists like volume normalization. With normalization, the relative volume of the tracks played do not jump around but remain fairly constant.
If you use the volume control with each new track, you are using volume normalization - it just is human controlled rather than automatic.
I think this is basically the implementation of the European Broadcast Union R128 protocol. I’ve used this, built in option within JRiver, for parties. It works well for a party but does end up lowering the volume for 90% of songs. The obvious solution is to turn the volume up.
I never use it for normal listening. I’m somewhat convinced it doesn’t sound as good on my system. However, everything I’ve read about R128 refutes my listening impression. I would need to do more comparative listening to come to my own solid opinion. The only problem with that is I hate comparative listening: too much like work and not enjoyment. Life is short.
Many pro R128 advocates insist it can eliminate the benefits of compressing music (loudest song on the radio or streaming playlist). If it just accomplishes that goal and leads to more dynamicly recorded modern music, I would be thrilled.
There is a lot to be said in favor of EBU R128. As it measures average loudness over time, a heavily compressed track will be played back softer as it exhibits high average loudness and, thus, needs to have its volume turned down to match the loudness of an acoustic jazz track, for example, which has lower average loudness.
In this sense, heavily compressed tracks are penalized.
Of course, this is also what any form of RMS normalization accomplishes.
99% of my listening is via playlists i create - been doing it since the days of tapes. I’ve found volume leveling in Roon and Tidal to be less than satisfactory to my ears but if Tidal have made some changes i’ll check it out. In itunes i use iVolume to level my library and i find it really useful. you can adjust by track, album or separate albums and apply different adjustments (than those applied to the rest of your library).