So I’ve had Google Analytics enabled on Theseus for a month now and I’ve got some stats now I can react to, and I have.
This graph show the amount visits per day since I installed analytics:

The spikes there occur when I update the game. I usually update in a weeks interval so I can get the game into the “Just In” list. Update more often than that and your app will simply just not show up there.
So now that I have some data I started to mine. I can see stuff like which is the most replayed (i.e. most difficult) level for old players (recurring visitors) and new players (new visitors). Good stuff; but I know that some levels are going to be harder than other.
I started digging into the stats and turned up with this:

The segment is new visitors and it measures how long they spend inside the game. While the average time spent in the game was something like 20minutes, over 44% left it under 10 seconds. Yikes!
Now some people may just start up a game and quit to try it out later, but hardly 44%. I’m doing something wrong here.
My reaction was to create a big “Start Playing!” button on the title screen which takes you directly to the first unsolved level. Tutorial 1 for new players.
I released it a couple of days ago and here are the same stats for the last couple of days:

Down to 37%. I’m not a statistics guy but I guess that’s outside the margin of error.
I guess that it’ll get even lower for regular days when the game hasn’t been in the “Just In” list.
This way of improving your product is a whole lot more effective than speculating what to do next. Feedback loops are important.
That is interesting. Can you break up the 44 % further somehow? For example by display resolution – maybe the game looks really ugly in some resolution?
Google Analytics doesn’t support it natively, but I guess you could provide extra data with the events you send to analytics and do custom analytics based on this. I think Flurry might have this out of the box. It was specifically made for mobile app analytics.
Google Analytics for Android is essentially just a wrapper for the web analytics and everything is presented in terms of a website. So e.g. loading a level is translated into a “page view”.
I might have chosen Flurry if I had known about it before. I still might, but I haven’t looked into it in depth yet.