In the latest version of our barcode scanning app for Windows Phone we added a very basic notification and analytic system that allows us to send short messages to certain user groups (Trial, Paid, Region, etc) as well as track how each group is using the application. This is all done anonymously by generating one web request per day to see if there are any new messages to display and including basic use information as part of the request URL (similar to how Google Analytics work). Eventually I would like to polish the code a bit and release it for others to use, but for now it is a work in progress that is starting to generate some interesting use metrics like the chart on the right, which shows that over 75% of users have scanned less than 5 barcodes in total!
This week I finally got around to using Log Parser 2.2 to filter through the log files and create some meaningful data. Unfortunately about half of the total aggregate downloads occurred before this system was in place, so we are missing some data, but from the data we do have we can find interesting results like the fact that around 66% of users are never seen again after the first day, meaning they never opened the app again after the initial 24 hours. That means that out of all the downloads listed on the AppHub reports (about 7,000 in our case) only 1/3 of them used the app more than once.
I’ve heard this reported before, but as I started digging deeper into the stats it became more and more clear that you only have once shot at impressing users and turn them into avid fans. Our app is a bit unique in that it is a tool used for a specific purpose that probably doesn’t need to be opened ever day or even every week, but with about two months worth of data collected it is clear that there are a lot of users that download the app, open it, and then for one reason or another never return.
This was most evident when we started running the numbers for daily active users. Our app has a pretty liberal trial policy (Unlimited free trial with no ads, just a nag screen), so I was not too surprised that the trial users out number the paid users, but what surprised me the most was seeing the daily active user breakdown of Paid vs Trial and New (aka initial app load) vs Returning. Below you will find a chart that outlines this data very well:
From this we can find that:
- Our daily user count has about doubled from around 60 a day to around 120 active users a day. This is lower than I would like, but not bad for a paid app with lots of free competitors.
- Trial users out number Paid users by around 10 to 1. The top two segments for groups A and B (Light blue and purple) show that most of our daily users are either new trial users or returning trial users. I might be able to increase conversions by limiting the trial version more, but my best bet would be to improve the initial experience to entice more of the new users to stick around.
- The New – Paid group is a bit larger then I would have expected at around 5 to 10 users a day, but this might also be due to paid users re-installing the app after reloading their phone. We track users based on an installation timestamp, so if you uninstall the app and then reinstall it, you are essential a New user.
- Since there are about 2,000-3,000 downloads that occurred before we started tracking users, we called these Legacy users and tracked when they first upgraded and opened the app. You can see a spike when the new software was released, and then a steady stream of a few legacy users each day. It is a bit surprising that there are users that have not opened the app in almost two months, but this just shows that our app is not a daily use app.
I’d like to dig into the data a bit more and find the average timespan between app loads for each type of user, but so far this system has shown that we really need to polish our first run experience to try and capture more of the initial users.