[retention] cohort retention chart and retention graph looks very different

  • 22 October 2021
  • 3 replies
  • 62 views

Userlevel 1
Badge +1

Hi 

thanks always:)

 

I wanted to ask you about the retention chart,

mainly about the differences in data

  1. difference btw the retention graph & cohort raw data within the amplitude
  2. difference btw the amplitude retention graph & app store connect

I made retention analysis chart as below

https://analytics.amplitude.com/Dreamfora0205/chart/new/jl350oa

(any event to any event, all users, strict calendar dates, N-Day retention)

It says 23.2% of users are returned on day1

 

However, when I downloaded the raw by eport csv,

day by day it seems like the retention is higher than 23%.

 

For instance on Oct19,

Day0 user were 619, Day1 user were 180 → 29%

so I added up all the users for Day0 from sep22nd to Oct 21st → 18215 users

                       all the users for Day1 from Sep22nd to Oct 21st → 2754 users

then the average day1 retention should be 2754/18215 = 26%

but, in the graph it says 23.2%

I guess it can have some differences because there’re some users might overlap.

 

Also, this retention is very different from what we can check in app store connect.

In the app store connect, 

the Day1 retention average is 32%, 

and for some specific days the cohorts are retained more than 40%…

so this looks very different from Amplitude which says 20% beginning.

 

Can you let me know why there’re difference for in between 1st one and 2nd one?

 

thanks in advance!

 

with best regards, Flor Cho

 

 

 

 

 


3 replies

Userlevel 1
Badge +1

Hi…

I would like to know the reason why the graph and cohort chart of retention looks different.

Can you help to let me know about this topic more in detal?

 

thanks!

Userlevel 5
Badge +2

The difference between the overall retention % and day-to-day retention figures is caused by the way Any Event to Any Event retention works on a daily level for returning users. Essentially a user who returns to your application every day would be counted as retained for every single day. This drives up the retention rates for individual days, but for overall retention figures this person is still only counted once.

You can see this in action if you go to your chart in Amplitude, scroll down to the table below the chart and click on the small arrow to open up the day-to-day view. If you compare the total number of users on the top row of this table to the daily number of users, you’ll see that the day-to-day numbers quickly grow larger than your total number of users for the time period, so you cannot just sum up the daily figures to get the total numbers in a chart like this.

If you started your chart from New User event, for example, which only happens once for each user, the situation would be different.

AppStore Connect gives you different numbers partly because it’s based on those users who opt-in to send statistics to AppStore, instead of all users, partly because it does not calculate the overall retention the same way Amplitude does.

I hope this makes sense and clarifies your situation, @flor.cho

Userlevel 1
Badge +1

@MikkoKarvonen 

 

Thanks a lot. 

so the problem was more about any event to any event,

because the starting point can happen repetitively.

thanks for your kind explanation.

 

About the difference between the app store connect,

Can you let me know the difference in btw the ways how the overal retention is calculated in Amplitude and App store connect?

 

Thanks

with best regards, Flor Cho

Reply