Cohort and retention dashboard

1. In AppsFlyer, from the side menu, select Analyze  > Cohort & Retention.
   The Cohort dashboard opens. The granularity and conversion date range display.
   
2. Select the cohort period granularity:
   • Hours
   • Days
   • Weeks
   • Months
3. Set the attribution period (user acquisition or retargeting engagement). Data relating to the most recent dates can result in partial data indicated by dashes in charts and tables.
4. Click Edit Cohort.
   The Cohort setup window opens.
   

   

5. Complete the following:
   • Cohort view type:
     • User acquisition.
     • Retargeting
     • Unified: Includes the in-app events of both user attribution and retargeting attributions. Note: Retargeting events take precedence over user acquisition events meaning the report contains events where is_primary_attribution=true. 
   • Trend type:
     • User lifetime performance. 
     • KPI by attribution time.
   • Users of: Select 1 or more apps. All means the top 100 apps in your account. Note: Agencies can only select one app.
   • Grouped by: Select one or more dimensions [Default: Media Source]. 
   • Minimum cohort size: Set the minimum cohort size. This avoids cluttering your report with insignificant (small) cohorts. [Default=1]
   • Optional fields:
     • Filtered by: Select one or more dimensions to filter by. There are no default filters.
6. Click Apply changes.
7. Select the metric (KPI) you want to perform cohort analysis on. For in-app events, you select the in-app event and metric. 
8. Select the cohort periods you want to view data for. Meaning which hours, days, weeks, or months post-conversion. The number of cohort periods you can select depends on your selected trend type.
9. Select whether to include partial data or to only show complete data (data for cohort periods that already ended). Learn more about partial data vs. complete data.
10. Select whether to show Cumulative or On day (hour/day/week/month) data:
    • Cumulative: The cumulative or running total of the metric. Meaning Cohort 3 shows totals for Cohort periods 0-3.
    • On day: The value of the metric for that day only. Meaning Cohort 3 shows the metrics of Cohort period 3 only; not counting what happened in the periods before.
11. Select whether to display the Heat map. 
12. Select whether to view the data as a table or chart. If you select table, you can select how many rows of data to see, either 25, 50, 100, or 200.
13. [Optional] Managing saved cohort reports.

• My reports enable you to save the current dashboard settings for future use. 
• Select a saved report to display the most current cohort data available.
• You can save one or more reports.
• Only you can access your My reports. 

• Cohort reports can be exported to CSV files.
• Columns names having a partial suffix is the equivalent of dashes that display on charts and tables in the user interface.

To export cohort reports to CSV files:

• In the upper-right side of the Cohort reports window, click (export).
  The cohort report CSV file is downloaded.

Cohort periods work as follows: The conversion period is referred to as 0 (Hour 0, Day 0, Week 0, or Month 0). The next period after the conversion is referred to as 1 (Hour 1, Day 1, Week 1, or Month 1), and so on.

The number of periods available depends on the period granularity selected:

• For hour granularity the maximum number of post-conversion hours is 72.
• For day granularity the maximum number of post-conversion days is 180. 
• For week granularity the maximum number of post-conversion weeks is 156
• For month granularity the maximum number of post-conversion months is 36

Note: In AppsFlyer a cohort period doesn't take into account the specific install timestamp. Rather, cohort hours are rounded down to the nearest hour, and cohort days, weeks, and months are based on the calendar day, week, or month. This may cause discrepancies when comparing AppsFlyer cohort data to the cohort data of other networks, where all cohort periods are determined by the specific install timestamp (meaning an hour is 60 minutes post-install timestamp, a day is the 24-hour post-install period after the install timestamp, etc.).

There are 3 cohort view types:

• Unified: 
  • The combined performance of user acquisition and retargeting campaigns (including cost and revenue).
  • If an event is attributed to both conversion types, only the retargeting event is included. Meaning, the report is generated using the actual last touch media source. If you are familiar with raw data, it means events where is_primary_attribution=true. 
• User acquisition: Performance data from installs (reinstalls excluded unless specially enabled)
• Retargeting: Performance data from re-attributions or re-engagements

There are 2 cohort trend types:

• User lifetime performance
• KPI by attribution time

User lifetime performance

• Each column in the table or point in the chart represents another cohort period (hour/day/week/month). The conversion period is referred to as cohort period 0. The first post-conversion period is cohort period 1 and so on. This enables you to compare user performance post-conversion, starting with the conversion period.
• You can select which cohort periods to get data for, up to the maximum for each granularity option (hour/day/week/month).

KPI by attribution time

• Each column in the table or point in the chart represents a cohort period (hours/days/weeks/months) on a period of the selected date range. This enables you to evaluate campaign performance over time and to view campaign KPIs relative to the conversion dates.
• The maximum number of cohort periods you can select to get data for is 5.
• If you select to view data as a chart instead of as a table, the chart displays only 1 cohort period (the lowest period selected).

Example: User lifetime performance trend vs. KPI by attribution time trend

In this example, the same raw data is used to generate the two cohort trend types. The raw data table contains attribution events related to different users A, B, C, and D. The users engage with an ad, install the app, and post-conversion make one or more purchases. Cohort day 0 is the install day.

For a given day in the conversion data range, if the number of days that have elapsed since the conversion date is less than the cohort day, that cohort day contains partial data. Meaning partial data occurs on any day where the cohort day > current date-conversion date.

Use the partial/complete data control to select the data display mode:

• [Default] Partial:
  • Attribution data updates continuously in real-time, including revenue.
  • Cost data updates 6 times a day on average every four hours.
  • Partial data (indicated by dashes in the table and charts) is included.
• Complete: Data is updated daily and available a few minutes after midnight. Data of partial days doesn't display. 

• The advertiser needs to grant the agency permission to use Cohort. 
  • To grant permission to an agency, the advertiser need to:
    1. In AppsFlyer, from the side menu, select Collaborate > Active Integrations.
    2. Select the agency.
    3. Go to the permissions tab, enable Allow access to your retention and cohort reports.
• Agencies are limited to accessing data of traffic generated by the agency. They can't see data related to organic traffic or non-organic traffic generated elsewhere.

Revenue sum by geo

Chart: Cumulative revenue per user

Table: Revenue per day

In-app event conversion

An advertiser defines the in-app event af_purchase. The cohort table displays the percentage of app users in the cohort who made a purchase. We can see on day 10 for instance that 36.27% of the cohort users attributed to the Facebook Ads media source performed a purchase.