“People who don’t upgrade their OS are also the kind of people who don’t buy apps”
This rings true to most developers. It likely depends on the type of customers you have (and thus the type of app), but the corrolary to that observation is that you should always lean towards dropping support for older OS versions. Brent Simmons made the case for OS X applications, but this should be applicable to iOS apps as well.
- Group A: all the users of his app
- Group B: the last 1000 users who purchased his app
He found that the iOS distributions for these two groups are basically the same. His conclusion is that there is no correlation between purchasing his app and upgrading the OS.
I do not understand the reasoning here, as I fail to see how the iOS distribution between those 2 groups of users could be different. Both groups are purchasers of his app, since this is a paid app with no in-app purchases. I don’t see why the purchasing and upgrade habits of the users who purchased it last year would be different from the habits of more recent users.
It is possible I am misunderstanding David Smith’s data, but to me, the more relevant comparison would be between these two groups:
- Group A: all the users of his recipe app
- Group C: all the iPhone/iPad/iPod Touch users
The iOS distribution for all iOS devices is published by Apple. If anything, it underestimates older iOSes, since it relies on App Store analytics from the previous 7 days. When plotting Apple’s numbers (from February 23rd) against David Smith’s numbers, one gets a very different picture. While 18% of all iOS devices are still on iOS 5 or iOS 6, only 10% of the users of his app are. This means that iOS 7 users are twice as likely to purchase his app compared to the slow updaters still on older OSes (the exact ratio is: 90 / 82 x 18 / 10 ~ 1.98). The ratio between slow and fast updaters is likely even higher, since some iOS 7 users are in fact slow updaters who just acquired a new iOS device.
I conclude that Brent’s conjecture is in fact supported by David’s data.