Those stats you're linking are extremely misleading and don't really reflect demographic of actual app users. Pretty much all our apps across the board have higher Lollipop and KitKat usage.
Which makes sense - what you linked is aggregate number of all users on the Play store across the world. Users that actually actively engage in their phones and western markets are noticably more up to date. Ignore newer API versions at your own peril.
I think it's more likely you're seeing a survivor bias. If your app runs well on Lollipop, it's very likely gigantic and slow as molasses on older phones due to the compatibility shim bloat.
> [..] demographic of actual app users [..] Users that actually actively engage in their phones and western markets are noticably more up to date.
Your post seem predicated on assumptions which aren't supported:
- That people with "better engagement" are more profitable to app developers (which might be the reverse of true if you aren't on a freemium modal).
- That non-Westerners don't buy apps at all or that we simply shouldn't care. .
No, no, I did not want to imply any of that. Just that:
- For pretty much any of our apps the actual breakdown of Android versions is nothing like what the Play store shows.
- The amount of new Androids is significantly higher (we actually have an app with 57% 5.x and 80% on 4.4 or newer - Samsung updates have been significant for it) and varies greatly
- The users with newer devices and newer Android versions are usually more engaged, use apps more and also advertise them more which makes for a good network effect.
- We have noticed a major shift to apps with Material design, arrogant Apple marketing aside, there's alot of users on Android that DO care about design, UX and those tend to have rather newish phones.
Depending on your app you should of course also care for people on worse phones. If you're providing a public service app (e.g. showing public transport schedules), you should probably target as low as 2.3 (even though, perhaps actually building 2 APKs for that usecase makes sense). If you're building a modern 3D game, you should probably target API 19+ since users with older devices probably won't be able to run it anyway. But in any case, DO YOUR RESEARCH, because the numbers vary greatly across demographics, app categories and regions.
1. Dashboard stats are global and in general show very different numbers when compared to the users of most modern applications, at least in the western world. It probably overrepresents cheap, old devices from China and India who, in general, don't install or buy as many apps.
2. As bad as Android adoption normally is, in general L adoption has been better or at least on par with that of previous versions. Check the last chart: http://www.bidouille.org/misc/androidcharts Lollipop release date is actually wrong there (makes the chart more optimistic), but by a month only.
Based on https://developer.android.com/about/dashboards/index.html though, L is behind the pace :(