MPKNOWLEDGEDROP

How are users moving from different states of engagement?

  • 20 May 2020
  • 1 reply
  • 695 views
How are users moving from different states of engagement?
Userlevel 2
Badge +1
  • Mixpanel Customer Success
  • 24 replies

Michael wrote last week about how to create engagement buckets. That’s a fundamental start, but the next question analysts invariably ask then is: how do we track how users move from one state to another? In other words, how do we track if users are becoming more engaged, or less? And more crucially then, how can we encourage movement between these engagement states?

 

Let’s continue with Michael’s example using the music finder. Suppose, we’re interested in tracking how users move from the Core cohort (play a song 1x a week) to a Power cohort (play a song at least 2x a week). 

 

We can construct the following cohort in the Cohort report to track the trend:

Cvzm2Pd-PeehpAvM-HNHcQcJlNwuuy54e7_M-s1CamnAcKSfbkdxmB4FygOG7IfaTzZ18UnfM7-ua6SABlOb8id5Uw4Ec4dFj2sManqrtrTBIHfW78Xh8BC23q7lApj2XiryJUlL

 

We can then visualise this cohort in the Insights report:

IeYZs--JMEcEAyqmTqODAHvAc5CIzb6sl4h0mOfX5pFrLfKVufkVbhpO9C0tBNm7eEVh6bMV0a2UgzpLw0Tz6fCrS28leNSa4RFY__IKYqGW_FkDHoKMmBBxnapmTivylaUB1zie

 

Now that we can track how users are changing from Core to Power, we can try to come up with different hypotheses of the actions we can do to grow that cohort over time. This is where the Signal report comes in handy. We can construct the following query:

PiK2Qe32GByB1r2nx4TBRVpfPMN4BUYbxv4105Ud28P-94g4x939AmIGovh6BdQ4UvL7Lx78ADy0ntS9b-7ol5poyentAEf1c-xNokQvk7QtlWGH6EC3vDHTzU1pmZJK6-HVXFMY

 

 

From the above report, we see the following two events that we might be able to take action on:

  1. Users who have used Search at least twice in the last quarter 

  2. Users who have followed a playlist at least 3 times within 7 days

 

Two possible actions we might undertake then, to try to increase the Core to Power cohort:

  1. Optimise the Search tool - could we put it in a more prominent position, or find other ways to encourage users to use the tool?

  2. Could we recommend playlists to users to follow? Or place the ‘follow playlist’ button in a more prominent location?

 

Then, as we implement these actions, we can track at a high level, if they have caused our cohort size to increase. Of course, there’s a more precise way (using the Impact report) we can measure if these actions do indeed contribute to the increase in the cohort size, independent of extraneous variables, but that’s a topic for another day!


1 reply

Peishan, thank you for the useful article! I look forward to the Impact report example!

Reply