Another thing is overlooked, which is strange since these companies hire so many statisticians and much about their work is about understanding populations and

Another thing is overlooked, which is strange since these companies hire so many statisticians and much about their work is about understanding populations and individual preferences. When you have huge populations (2B) you're outliers are going to be similarly huge. The meaning and usefulness of means and medians over large and/or disparate populations loses its meaning. There are tons of distinguishable subpopulations (Dutch) which are pointless to lump in with Americans. Etc etc etc. You'd think someone would be familiar with subpopulations and the limitations of treating 2B users as a normal distribution, but yet that seems yet to be discovered. Ridiculous.