I've seen interviews with certain games companies where they talk about getting stats back from their users, particularly with games on facebook. This data allows the companies to monitor and adjust the difficulty of levels, see what's keeping people engaged etc. I am just wondering exactly what sort of data would they be collecting? How common is this practice and would developers of Android and iOS games be involved in similar practices?
It's most common for developers to track simple usage. That is, how often people play a game, how long they play it per sitting, how often they return to it, and which features of the game they use the most.
For games in which the player character can die, they often track where deaths happen. For puzzle games, they track puzzles which people skip, or simply stop playing at. All of this is aimed at identifying awkward spots in the difficulty curve.
I have personally worked on games where we tracked almost every choice that players made during the game. We didn't do this in order to track individual users, but rather to notice the larger trends. As an example, one game I worked on was a game where at the start of the game, the player was given a selection of five cars to drive. They could pick any those five cars, for free. And we tracked which of those five cars users selected. What we learned from that tracking helped inform our user interface designs for future games.
You can safely assume that every game which has any online component will be tracking this sort of data (this includes everything using GameCenter or Crystal or OpenFeint or etc). You can also assume that every game from a major publisher/developer will be tracking this sort of data, even if they don't use one of those services.
The question really isn't "what stats to track", but should be thought about in terms of "what problems might come up".
As an example:
At least in the Zynga-style Facebook world, the goal of stat tracking is to improve your game over time so people stick around longer and buy more things. Stat tracking is a means to an end. If you don't have good visualization software or ways of comparing correlated data, then just grabbing raw stats is kind of meaningless. Likewise if you don't have directed questions you're trying to get objective answers to, your stats don't really have any context.
Funnel stats in particular are a good one to look for. If you notice that 10% of your users don't finish a particular tutorial step, then maybe that part of it needs to have some design time put into it. If you have a lot of people leaving before the game finishes loading then you know you need to get your file size down. That kind of thing.
Another popular thing to do to use stat tracking effectively is A:B testing. Let's say you want to make change X that should improve some metric Y you care about. For the majority of users, leave the game the same. Introduce the change to a small percentage of your users (but large enough to be statistically significant), and compare results. If your change X increases metric Y, then you start rolling it out to more and more people. If it doesn't, you revert it. This is useful to get tuning for costs right.
Ever wondered why there are so many useless, but progress-indicating ingame achievments like "played through the second level on any difficulty setting".
Steam Achievements for example pose a horrific opportunity for gamedevs to track their products usage. So for a real time answer just browse any achievements page for a game.