If you don’t know the word “cohort,” this is what the dictionary says:
Even though it could have been interesting to start mimicking military organizational elements from the Roman empire, the User Retention Cohorts that we recently introduced in Telerik Analytics are more related to definition (2) above. Although the report type does have the power of a small military unit…
It can all sound a bit complicated, and to some extent it is, but if we keep track of the intention with user retention cohorts, then it gets easier to grasp. Basically we wish to present data that will help you understand how loyal your users are. What type are they? Do they use your app once and never again? Do they use it multiple times every day and seemingly without any end? Where are they on this scale?
When we speak about users in Telerik Analytics we actually mean unique installations of a given software application running on a set of devices. This could, for example, be a given hybrid app running on a number of different mobile phones. Each running app instance would be what we call a user.
A user cohort is a subset of all users that you are monitoring with Telerik Analytics for your application. In other words if we look at all the users of one of your apps, then a user cohort is simply a subset of these users.
In this first release of User Retention Cohorts the shared characteristic for the cohorts is simply the time the application is started the first time. You could say, when the “user is born.” We have chosen to implement three different cohort types: day, week and month. These correspond to the timespans that will determine whether a given user is in one cohort or the other. We basically group users by the day they were born, or the week or the month. A cohort could be all users born on Aug 12th 2015, or all users born in January 2012.
Let’s look at an example. Below we have a number of users listed with the time of the first execution of a given app.
Initial execution (“born on”)
Fri, Aug 7, 10:30
Fri, Aug 7, 20:50
Sat, Aug 8, 14:21
Mon, Aug 10, 14:21
Mon, Aug 10, 17:01
Mon, Aug 31, 22:25
Tue, Sep 1, 07:44
The names of the individual cohorts (D1, W1, M1 etc.) are not relevant. They could be anything, but they show for each specific user (A to G) which cohorts the given user belongs to.
Each user is born once and thus belongs exactly to one day, one week and one month cohort. This never changes, like your birthday, birth-week and birth-month never change.
In the table above, you can see that users A, B and C all belong to week cohort W1. User A and B belong to day cohort D1, and C belongs to day cohort D2. So they are all “born” in the same week and A and B are born on the same day.
For the three cohort types we have decided that:
With this definition of cohorts we can talk precisely about subsets of users and start asking questions like: for the specific user month cohort that holds users born in April 2015, how many of these specific users used our application in May 2015, in June 2015 and all the way to, say, August 2015? In other words how many of the users in the given cohort were still “alive” in the following months?
This information is very important because it shows you how long an attention period you have from your users and how quickly or slowly they lose interest in your application. If you see a huge loss of users within a very short time, then clearly you have a challenge to improve your offering.
With the User Retention Cohorts you can choose to view day, week or month cohorts depending on the type of business you drive.
User engagement can be affected in many ways. Obviously the particular version of your application is an important aspect, but marketing campaigns, user reviews and even the particular time of the year or maybe even the time of the month or week can have an effect.
This is why we have made it very easy for you to compare cohorts and see how they “perform” over time. With this view you can gauge whether the steps you take to improve your business work or not.
These cohorts are one reason among a growing list to use Telerik Analytics, and we plan to improve the User Retention Cohort reports over time. We believe it can be relevant to define cohorts based on more shared characteristics. It could be the country the user lives in. It could be those users that have executed a specific feature in your app. It could be users that, on the other hand, haven’t performed a specific action. With such further analysis you would be able to understand segments of your users better and thus be able to take the best next step to improve their experience with your app.
You can read more here:
Eigil Rosager Poulsen is the Product Manager for Telerik Analytics. Having founded the application analytics company EQATEC, which was acquired by Telerik in 2013, he is passionate about helping application developers collect the data needed for better decision making. Eigil lives in Copenhagen with his three daughters.