The Cohorts screen allows deep analysis into your retention across time. To Cohort is to group and is frequently called ‘bucketing’, as in bucketing similar users together. Cohorting allows users who are similar in some way (date of signup, mobile device type, industry, number of purchases, etc) to allow their retention to be compared.
Cohorts is a powerful way to see how you are retaining users, how your retention changes between populations, and to track how your retention changes over time (such as between application versions). Cohorting in Appuri is fully on demand, thus you can run complex calculations from the Appuri interface.
Unlike other platforms, Appuri allows on-demand cohort views for any past events and event properties. This means cohorts are calculated in real time, thus there may be a several second delay from modifying a cohort to seeing the results populate.
Cohorts is only available for customers using Appuri’s Retention Management solution.
When you first load the Cohorts page, the default options are
Everyone cohort by
first event over
in the last 8 weeks. This is a great basic cohort.
First, this is including all users. You can apply a segment to limit this to a more specific audience, if you want.
Second, this is based on the last 8 weeks. This means two things. First, the oldest cohort will be from 8 weeks ago. Next, the users will be grouped into 8 groups, one for each week. By contrast, the
in the last 30 dayswill have the oldest cohort be from 30 days ago, and users will be grouped into 30 groups, one for each day. In the first example, the unit of time is weeks, whereas it is days in the second example.
If your typical customer lifespan is measured in days, you may want to change the cohort to be
in the last 14 days. If your customer lifespan is measured in weeks or months, you may want to change the cohort to be
in the last 12 weeks
- Third, the colored table shows the percentage of users who retained in the next unit of time after their signup. If this is weeks, the first box represents what percent of users returned between the 8th day and the 14th day after the start of the cohort. Note a user only needs to return during one of these days to be counted. Returning on several days, or several times a day, will not impact this percent as only their first return is counted.
Additionally, note that each row is of a different length. This is because the first row is the oldest cohort. They started the soonest, and they have been around the longest. Thus, they have more weeks of data available. The next row joined the next week, thus they have one less cell than the older cohort, and so on.
The first time a user shows up, during their first week, there is 100% retention for the cohort.
This will always be the case when cohorting by
first event. Thus, Appuri does not display the first week which would always be 100%. Instead, Appuri shows the total number of users who belong to this cohort.
There are additional options for using the Cohorts screen. This section dives into advanced usage.
On the top right of the cohort screen, you can change the cohort display. There are two options here.
First, you can change the cohort chart type. This makes each Cohort (each row) into a line on a line chart. Thus, it is easy to compare where a given cohort falls in comparison to all others on the given unit of time.
By default, Cohorts display as the percentage of users who were retained in the selected time range. On the top right corner, you can change this to the absolute number retained, by clicking on the
# symbol. This changes the percentage of users retained to the counts of users retained.
On the left side of the cohort screen, you can modify the cohort definitions.
First, you can cohort by different events. The default is
_first_event_at, which is an event created by Appuri. If you have an event to represent an event you are interested in seeing, you can select it here. Interesting options may include the date a user first signed up for a subscription or first made a purchase. Other interesting events to cohort on may be when a user submitted a support request, first invited a friend or collaborator, or first engaged with a specific feature.
You can select any date in the past to view a snapshot of retention at that point in time. When using this feature, the retention graph will be filled out, thus each row will have the full number of columns if it is far enough in the past.
Selecting this option will change how users are grouped. Rather than grouping by time (default) you can cohort based on a specific event property. This could include:
- Which campaign or referral the user signed up from (Facebook campaign, Google campaign, etc)
- Which platform or operating system they are using (iOS, Android, etc)
- Which browser they are using (Chrome, Firefox, etc)
- If they have seen an error such as a crash or failure message
- Retention for users by version number (version 1.2 vs 1.3 vs 1.5)
In all of these cases, users from all time will be grouped by which meeting the condition, thus each row on the results would be their respective values, showing that, for example, All Android users have a 68% day 30 retention, while iOs users have a 52% day 30 retention.
Because these cohorts are grouping users for all time, if you have a significant amount of historical data you may want to create and apply a Segment such as "Users who signed up in the last 6 months" to limit who is included.