Alternately, “If you don’t assume that all usage is fungible, the data tells a different story”. There are a few apps - music/movies/ebooks and Facebook - which get a disproportionately large amount of activity. If you’re not competing with them, the global numbers are going to be skewed.

This is especially true when you think about tasks rather than screen time. If I’m Facebook or Spotify selling ads, I want people glued to the app for a long time. If I’m selling things, giving information or customer service, etc. the story is completely different: someone placing an order faster is actually a win.

EDIT: My experience is colored by being in the cultural heritage space where the flurry of “we’re in the App Store!!!” entries met the hard reality of every engagement metric being much lower than web sites. There is huge variation hidden in global averages.

Well, "90% of user time on devices is in native apps, 10% web" does paint a rather different picture than "70% of user time is in social media and messaging apps, 10% media apps like Netflix or Spotify, 5% mobile games, 5% all other apps including all the built-in ones other than the browser, 10% web" if you're not in social media, media, or games and trying to decide whether to go web or native.

Wait, are you saying that "90% of mobile time is spent in apps vs 10% in browsers" tells the story that "consumers vastly prefer native apps"? I don't think the connection is that strong.

Taking myself as an example, I may spend 90% of my time on my cell phone using the Messages app, the Phone app and a handful of apps that I installed a long time ago, but I'm much more likely to try out a new web app than I am to try out a new native app. Looking at my phone now, I see exactly one native app that I installed in 2020.

Based on my and my friend's preferences, we prefer mobile app for apps they will be using a lot and trust, like Facebook/Slack/Twitter/Email, and want to use the web for apps we rarely use or don't trust as much.

Op is arguing that based on our intuition the data aggregated at that level doesn't distinguish between this more complex story and the simple one of "people prefer apps".

Or

"It's often useful to take context into account when evaluating statistics"

If some of the data are outliers, they arguably should be removed.