A cohort is a group of users who have performed a particular group of events or share a common set of property values. You can use cohorts to analyze a specific group of users in the Analysis reports or to target that group of users through Messages.
You do not need to create user profiles to create a cohort. Cohorts can also include anonymous users. View a list of all of your project's saved cohorts by clicking on the Cohorts tab.
Create cohorts based on users or groups by clicking on Users at the top of the page.
Create a New Cohort In Users
Create a cohort in the Users tab under Explore by building a cohort to search your users.
You can build a cohort based on events, event properties, user properties, and even other cohorts. You can also create a cohort that combines multiple existing cohorts.
First, select an event, user profile property, or cohort from the first drop-down menu. Each set of criteria that you select is considered a group.
If you define multiple groups, click on the AND button to tailor your logic to fit your use case. When you click on AND, the option switches to OR, which gives you greater flexibility in targeting.
Circular logic in cohorts creates an error. Here are a couple examples of circularity that break a cohort:
- Cohort A = Power Users
Cohort B = Churn Risk Users
Cohort C = High-Paying Users
Cohort D = Cohort C OR Cohort A
Then change Cohort A = Cohort D
This will result in “Power Users = Power Users OR High-Paying Users”.
- Cohort B = Users NOT IN Cohort A
Cohort A = Cohort B
This will result in “Cohort B = Users NOT IN Cohort B”.
Filtering by Events
If you select an event, you can choose how frequently that event is performed or not performed by the users you are targeting.
If you filter for events that were performed more than once, you can specify whether users need to perform it over a period of time, or over a period of time on unique days.
When you select an event filter, you can add more specific event property filters by clicking the + Filter button.
You can add more conditions to this targeting event by clicking the + Condition button in the top right.
Use the OR, AND, THEN logic options alongside the ANY and ALL options to tailor your cohort to a specific group of users.
Filtering by Property
If you select a user profile property as a filter, the options available in the filter will vary depending on the data type of the property (string, number, boolean, or date).
You can select user profile property filters by clicking the + Filter button.
You can also add one or more conditions to the property filter by clicking the + Condition button at the top right of the group. This serves as a second property filter.
You can then select whether you want your targeted users to have both the targeting property and the condition, or one or the other.
Filtering by Cohort
If you select a cohort, you can select whether you want to target users in or not in the cohort.
You can select additional user profile property filters for that cohort by clicking the + Filter button.
This can be used to build cohorts from pre-existing cohorts. Note that if you do so, you will not be able to delete the dependency cohort until you delete the original cohort.
For example, you can create a new cohort A, which combines cohorts B and C and some event properties, but you will not be able to delete cohorts B or C.
Learn more about deleting cohorts here.
Create Cohorts (new experience in beta)
We are rolling out an upgraded experience for building cohorts - please reach out to email@example.com if you'd like early access. With this change, you will be able to use property aggregations and 1st time filters to build cohorts and answer questions like:
- Show me users that did Purchases of a cumulative sum of $50 in the last 7 days (as opposed to what’s supported today, which is “show me users that have done A purchase of more than $50 in the last 7 days")
- Show me users that watched more than 2 unique videos in the last 30 days
- Show me users that watched a video in 4 unique weeks in the past 3 months
- Show me the users that did their 1st ever purchase in the last 30 days
As mentioned earlier, a cohort is a group of users that have performed a particular group of events or share a common set of property values, so you can build a cohort starting with the "Add Filter' button and filter down by events, event properties, user properties, and even other cohorts.
Filter by events
If you want to build a cohort of users that performed a particular group of events, you can start by hitting the filter button and select the event name, following up with the frequency and the time range:
You can also add event property filters to this to make your targeting more precise:
You can also add a 1st time filter to an event to get a group of users that did that particular event for the very first time ever:
What this example shows is that there were 16,065 users that had watched their 1st ever video in the last 30 days.
You aren't just limited to building cohorts based on event counts - you can now also build cohorts based on property aggregations. Simply click on "Total" -> "Aggregate properties" to then select what kind of aggregation you'd like to do.
Here's an example of creating a cohort of users that watched a cumulative of over 300 mins of video in the previous 30 days:
Here's an example of creating a cohort of users that watched videos of an average watch time greater than 15 mins in the previous 30 days:
Here's an example of creating a cohort of users that purchased items across 3 unique categories in the last 30 days:
Creating a cohort based on users doing events on "unique days" hasn't gone away - in fact, users now have the ability to create cohorts of users that did events on X "unique hours", "unique weeks", "unique months", "unique quarters", "unique year", "unique hour of day" and "unique day of week" and users can now get to that using the property aggregations.
Here's an example of creating a cohort of users that watched a video on at least 3 unique days in the previous 30 days:
Here's an example of creating a cohort of users that made a purchase on 4 or more unique weeks in the last 3 months:
See how easy Lifecycle analysis becomes with this new cohorts builder:
Filter by properties
You can also filter by user profile properties and lookup profile properties by hitting the "Add Filter" button, here's an example of filtering "All users" that are above the age (user profile property) of 60:
Here's an example of filtering "All users" that are from Asia (where Country.Region is a lookup profile property):
Filter by cohorts
You can also define a cohort by inheriting the conditions from a cohort that has already been defined (thereby not needing to define the cohort again!) and here's an example of creating a cohort of users that are part of both "Active video watchers" and "New users":
You can also mix and match event, property and cohort filters to define a single cohort, and here's an example:
If you want to add multiple conditions onto a single group, you can still do that by hitting filter and adding more conditions, you can then change how the different conditions are applied in that group by selecting "and":
Here's an example of creating a cohort of users that did at least 1 purchase OR did 2 or more video watches AND created 1+ playlists:
You can also use "and then" condition that was available earlier by selecting "then did". Here's an example of users that did 1+ purchases and THEN DID 2+ video watches and THEN DID 1+ create playlists in the last 30 days:
Save a Cohort
When you finish defining the group filters for your cohort, click Save. You must give your cohort a name if you did not do so in the group filter screen. You can add a cohort description to give your team more context about the group of users in your cohort. You can choose to just "Save" the cohort and keep that cohort private, or save the cohort and share that cohort with your colleagues in the same flow (as shown below):
If you already hit "Save" without sharing that cohort with anyone - not to worry, as you can always hit the "Share" button after saving:
Update a Cohort
You can update cohorts and any changes you make to your base cohort will reflect in other cohorts that depend on it. For example if a Churn Risk Users cohort (A) is defined as “users who have not done any event in the last 30 days”, you can create a dependent cohort called Churn Risk Users in San Francisco (B), which is “Users in Churn Risk cohort AND have the property ‘City’ = ‘San Francisco’”.
Now, if you make any changes to Cohort A, such as changing the date range to “users who have not done anything in the last 60 days”, then Cohort B will automatically update to use the new definition of Cohort A.
Create a New Cohort In Analysis
You can also create a new cohort directly from the Funnels and Retention reports. Learn more about these methods below:
Create Temporary Inline Cohorts
You have the option to create a temporary cohort in Insights, Funnels, and Dashboards when you choose to add a cohort, or filter and breakdown by a cohort. This cohort will only exist for the current query and does not persist throughout Mixpanel.
In the Insights report a temporary cohort can be created in both the query and the results of the query as a filter or breakdown.
In the Funnels report a temporary cohort can only be created in the results of your query as a filter or breakdown.
In Dashboards a temporary cohort can only be created to filter your dashboard.
To add a temporary cohort to your Insights query, click Add and select Cohort. Then, select Create cohort.
To add a temporary cohort as a filter or breakdown in your report, click either Filter or Breakdown and select Create cohort under the "Cohort" tab.