How to Analyze Traffic Funnels and Retention in Facebook Applications

In several months, Facebook’s event, f8, is going to launch its third hackathon. Every year I like to take a step back and ask, “what did I learn?” With Mixpanel, we have the opportunity to work with lots of companies and developers, helping them really understand what their analytics mean with regard to their applications. I am going to go over some of the general things that I think all developers can benefit from:

Visitor retention

Also called “cohort analysis,” understanding visitor retention is probably the most underrated yet most important metric all Facebook developers should be looking at. Visitor retention helps you understand how “sticky” your application is by being able to tell you the percent of users that came back to your application 1-6 weeks later.

In the early days of Facebook’s platform, lots of the applications with the highest amount of installs have now failed because their retention was extremely low. The only way to increase this is by giving really good reasons to users to come back again and use your application. Sadly, applications like Hug Me were likely only fun the first few uses.

Max Levchin, CEO of Slide — my former employer — has described what happened with applications early on as large hollow spheres where recently acquired “users” made up the outside. This was in part due the sheer number of users inviting each other and accepting invites/notifications. The ideal app is a sphere filled with users acquired previously as well–meaning returning users combined with new users which make up the outside.

This breaks down as follows:

< 20% (week over week) is poor and needs serious improvement.
< 35% (week over week) is pretty good > 45-65% (week over week) is excellent and ideal.

The reason games on Facebook really took off was large in part due to great retention in comparison to competing top applications. Games provided really great game mechanics and incentives to come back. The concept of “farming,” for example, has lots of replay value.

With retention, we generally see that really good applications have fairly consistent week over week retention. This means even 3 or 4 weeks later the retention is roughly the same as the number of users who came back just 1 week later.