In cohort analysis, we take a group of users with a shared conversion date (usually, users who installed an app on a specific date), and look at their behavior over time. For example, we can see:
- How much revenue did this specific group of users generate within 7 days of converting?
- How many of the users are still active by day 14?
One of the goals of cohort analysis is to separate user-engagement metrics from growth metrics. It lets us measure the behavior of existing users over time without it being masked by engagement on income that's coming in from brand new users.
Troubleshooting Cohort Data
We don't recommend comparing cohort analysis between different sources unless you can confirm that all the sources are using both the same attribution method and the same method for calculating cohorts. Any slight differences can cause the numbers to look different.
Note the following points in particular:
- Singular groups users into cohorts based on the install time (not the time they clicked on the ad).
- In Singular, once a device is attributed, all the events from the device are attributed to the same source. There is no expiration date for the attribution decision - it changes only if the device is re-attributed (see the Re-engagement FAQ for more information).
If you are comparing global cohort data (not broken down by attributed source), you may want to check the actual numbers for the event, i.e. how many times the event happened on the actual date.
In Singular, you can get these numbers by choosing "actual" for your cohort period in reporting, or by downloading event logs in the Export Logs page.
If there is a significant difference between Singular's numbers and your platform's numbers, this indicates an issue beyond cohort calculations.
To run a cohort report in the Reports page:
- Choose a date range. This will define your cohort(s). For example, if you choose November 1, you will see data for the cohort of users who installed the app on November 1. If you choose November 1-3, you will see data for 3 cohorts.
- Select dimensions as usual.
- Select one or more Cohort Metrics. These are the metrics that Singular can track in a cohort-based manner, including Revenue and custom events.
- Select one or more Cohort Periods. These define the number of days post-install for which you want to track the cohort's behavior.
For example, select the January 1 date, the Revenue cohort metric, and the 7 day cohort period to see how much revenue was generated by users who installed the app on January 1 in the 7 days after the install.
Note that cohort metrics are accumulated. If you select multiple cohort periods, the metrics for the longer periods will include the events and revenue generated in the shorter periods.
Therefore, for the following settings:
The resulting report may be:
This means that the cohort generated:
- $20 of revenue in the first 24 hours after the install
- Another $20 between 24 and 48 hours after the install, making a total of $40 for the first 2 days
- Another $15 between 48 hours and 7 days, making a total of $55 in the first 7 days.
If you choose a date range of more than 1 day and a Time Breakdown setting other than "Day", the metrics for all the daily cohorts are added up.
For example, the following report includes three cohorts: the group of users who installed on 2021-01-01, the users who installed on 2021-01-02, and those who installed on 2021-01-03. In this example, under the setting Time Breakdown, we chose Day.
If we select the breakdown All Time instead, the metrics for the daily cohorts are added up:
Cohort periods are the number of days for which the cohort metrics are calculated. If you select a cohort metric, you have to select at least one cohort period. You can select multiple periods.
|1d, 3d, 5d, 7d, 14d, 30d, 90d*, 360d*||A number of days for which to aggregate the data. "14d" is selected by default.
* Cohort periods of 90d and more are available only to customers of Singular's attribution service.
|Select the date Dec 1, 2019, the metric Revenue and the period 14d to see all the revenue garnered from users who installed the app on Dec 1, in the 14 days after installation.|
|LTV||Lifetime Value, i.e. aggregation of the data for the period starting in the conversion date and ending today.||Select the date Dec 1, 2019, the metric Revenue and the period LTV to see all the revenue garnered so far from users who installed the app on Dec 1.|
|Actual||Select this to get non-cohort data, i.e. aggregation based on the actual date in which the event occurred rather than based on cohort.||Select the date Dec 1, 2019, the metric Revenue and the period Actual to see how much revenue was gained from all the users of the app on the actual day of Dec 1.|
Singular calculates cohort events based on a 24-hour rolling window from each user's install time.
Sometimes you may want to see stats from the Cohort Metrics section, such as Revenue or ROI, without having them analyzed by cohort. Instead, you want to see the stats for the actual day in which the events occurred. In that case, you can select the "cohort period" Actual.
For example, if you select the date range 2021-01-01 - 2021-01-02, the metric Revenue, and the period Actual, you will see how much revenue was gained from all the app users on the actual dates 2021-01-01 - 2021-01-02, regardless of when they installed the app.
The default metrics available for cohort reporting in Singular are Revenue, ROI, and ARPU (Average Revenue per User).
You can also add your own custom cohort metrics by creating cohort events. This allows you to track any event that occurs in your app in a cohort-based manner, including registrations, video views, level-ups in a game, different types of purchases, etc.
To see which cohort metrics (default and custom) are currently configured in your account:
- In the web app, go to Reports and look under the Cohort Metrics section.
- If you are using the Singular API, query the Cohort Metrics endpoint.
To add a new cohort event:
- Make sure that the app reports the event to the attribution provider. If you're using Singular attribution, this means having the Singular SDK track the event.
- In the Singular web app, go to Settings > Events and create a cohort event. Map it to one or more in-app events coming in through the Singular SDK or third-party tracker.
After you've defined a cohort event, it becomes available in Singular reporting alongside the default cohort metrics.
Singular defines a re-engagement event as the user's opening of an app that they have already used in the past, following a click on a retargeting ad.
In-app events and revenue generated following a re-engagement event are attributed to the network and campaign that are responsible for the re-engagement (rather than to the source of the original install).
Therefore, the new events and revenue become part of a new cohort: the cohort of the re-engagement date.
For more information about re-engagement and retargeting in Singular, see the Re-engagement FAQ.