import { Callout } from 'nextra/components'

Cohorts: Group users by demographic and behavior

Cohorts are groups of users that share a certain set of properties or who perform a similar sequence of events. Mixpanel lets you visually define cohorts, view the list of users that comprise them, compare them in your analysis, and share them with the rest of your company.

Here are some examples of cohorts you can create in Mixpanel:

US Users: Users who are from the US
New Users: Users who signed up in the past month
Power Users: Users who used your app on 5 out of the last 7 days
Dropped-Off Users: Users who signed up, but did not come back the following week

Creating Cohorts

You can create cohorts via any of our report visualizations or explicitly using the cohort builder.

Creating a Cohort via a Report

Click any bar or point in an Insights, Funnels, or Retention report and select "View Users". The side panel will show the list of users that belong to that bar or point you selected. You can save that group of users as a cohort.

Creating a Cohort via the Cohort Builder

The Cohort Builder lets you define cohorts based on a precise set of conditions. These conditions are either of the form "Users who did <event> more/less than <some threshold>" or "Users where <profile property> is equal to/more/less than <value>". You can save these conditions as the definition for your Cohort.

Below are a few examples:

Filter by users who watched more than 10 videos in the last 7 days

Filter to only users who watched a video for the very first time in the past 30 days

Chaining cohorts together using the AND/OR operators

You can also access the builder directly from the query builder in any of our reports, by clicking Create Custom > Cohort. This will create a temporary cohort that only persists for your analysis session.

Note on cohorts whose filters contain user properties

Because user profile properties only store the most recent value, cohorts involving user profile properties will use the current value for those properties (even if the value changed over time).

For example, suppose the cohort’s filter criteria is: users where user[“City”] == “SF” and “Did Event: Order Ride 3 times in Last 7 days”.

Mixpanel computes the above over the last 30 days and then groups the set of users who have property user[“City”] == “SF” as of right now and intersect that with the daily cohort of users who did Order Ride 3 times in the last 7 days.

Creating a Cohort via Lexicon

Cohorts can also be created and managed in Lexicon. In Lexicon, you can see all your saved cohorts alongside your other saved definitions like metrics, custom events, and custom properties.

Creating a Cohort with Changing User Properties

With historical user or group properties, you can optionally Add Time and Measurement to specify the time-range over which to search for the property value

This brings up time selector options.

To go back to the default of using the latest value of the historical property, you can Remove Time and Measurement.

Creating a Cohort via a CSV

You can create a Cohort by uploading a CSV. This method is useful when grouping Mixpanel users who cannot be easily identified through specific behaviors or user properties. (e.g. when you have a list of distinct_ids to include in a Cohort and they were randomly selected)

We will use the CSV to update users in bulk by adding a user property that indicates their Cohort membership. Afterward, we will use this property to define the Cohort.

For example, if you have a list of 1,000 distinct_id values to group into a Cohort, create a CSV file with two columns: one containing the distinct_id values and another containing a user property value that identifies cohort membership (e.g. inCohort == "yes"). Exclude column headers from the CSV, as you will define them during the upload process in the CSV upload wizard within the UI.

After you create the CSV, go to the Users tab, click Add/Edit Profile, select Import from CSV, and upload your file.

Next, define the property names for each column in the CSV uploaded. Assign "$distinct_id" as the property name for the column containing the distinct_id values, and choose a property name for the column containing the user property value. Once you verify that the column assignment is correct, click Import Profiles.

If your CSV includes a `$distinct_id` value that does not match an existing user in your project, the system will create a new user with that distinct_id.

Finally, use the newly uploaded user property to define your Cohort.

Note that Cohorts created using this method are static; they will not update over time, as the associated user property remains unchanged.

Using Cohorts in Analysis

You can visualize how cohorts size over time, use cohorts to filter your reports, or compare how cohorts perform a metric.

Visualizing Cohort Size Over Time

In Insights, you can define a metric based on the size of a cohort over time. This is useful to understand how key subsets of your user base (like your Power Users) are trending over time. Select a cohort from the "Events and Cohorts" menu in the Insights report; this will generate a line chart that shows the size of the cohort over time. Each point is the size of the cohort as of the end of that time interval.

Filtering by a Cohort

You can use cohorts to filter any of your analysis by selecting a cohort from the filter menu. For example, if you're focused on Onboarding, you can filter all your analyses to the New Users cohort, which might be defined as users who signed up in the past 7 days.

Comparing Cohorts

You can compare the behavior of different cohorts by selecting a cohort from the breakdown menu. For example, you can use this to see how engagement compares for Power Users vs All Users.

When breaking down metrics by multiple cohorts, events are evaluated independently against each cohort. Consider an insights report that counts all events broken down by 3 cohorts, 'Power Users in North America', 'Power Users in Europe', and 'All Power Users'. An event belonging to a power user in North America, would contribute to the sum for buckets, In 'Power Users in North America', In 'All Power Users', and Not In 'Power Users in Europe'.

How Cohorts are Computed

**Note:** To ensure timely results from queries, newly imported or changed historical events timestamped more than 30 days ago might be part of your existing cohort analysis only after 12 hours from their import/change.

Cohorts are computed dynamically at the time that you use them in a query. Suppose you make a funnels report that filters down to the New Users cohort. Under the hood, Mixpanel does the following:

Run a query to compute the set of users who are in the cohort as of right now (the time that the query is run).
Run a query to compute the set of users who converted in the funnel, filtering down to the list produced in Step 1.

This has a few implications. First, it means every time you run a query that uses the cohort, the cohort is freshly computed vs being a static snapshot of users. Second, it means that the set of users who are in the cohort are users that are in the cohort as of right now; it's not a rolling window of users that have ever been in the cohort.

Cohorts and Time

A common question that customers want to answer is how certain user behavior impact downstream behavior. Should you use cohorts or funnels to answer the question of how many users watch a comedy video and then subscribe? Cohorts and filters will give you different answers.

For example, consider this report: https://mixpanel.com/s/4AuMRi

For the cohort on Jan 1, I'm looking at the users who watched a comedy video in the last 30 days before Dec 25 and who then subscribed over the same period of time. However, for the funnel on Dec 25, I'm looking at the users who watched a comedy video on Dec 25, and who have then also subscribed anytime within the next 30 day period from Dec 25. This means that if a user subsequently subscribes on Jan 3, the funnel chart can change to take that conversion into account. In this case, if you care about the conversion window, i.e. the amount of time that a user has in order to count as a conversion, using funnels would give you a more accurate answer.

Saving and Sharing Cohorts

Growth and Enterprise plan users can save an unlimited number of cohorts. Free plan users can create cohorts but cannot save them. See our [pricing page](https://mixpanel.com/pricing/) for more details.

Cohorts that you create are, by default, only visible to you. You can optionally share cohorts with specific other people or with your organization. This makes Cohorts that you create discoverable by others.

We recommend sharing a set of key cohorts with your organization, to ensure that you align on key definitions of user behavior that are important for your business (eg: New Users, Power Users). We also recommend adding descriptions to such cohorts, to share context with your team.

Exporting Cohorts

You can download the list of users in any Cohort as a CSV via the Cohort Builder. You can also push cohorts to 3rd-party destinations like Segment, Braze, or a Custom Webhook.

FAQ

How can I create cohorts based on dropped-off or retained users?

There are two ways to create such cohorts.

The first way would be to create a Funnels report and click the respective segments in order to find the retained/dropped-off users. You can then click “View Users”, followed by “Create Cohort”. Here is a Funnels report to play with: https://mixpanel.com/s/8JY6o

The second way would be to use the Users page to create cohorts. For example, if you would like to find users who “watch video” and then “like video”. You can create the cohort in the Users page. Here is the cohort for reference: https://mixpanel.com/project/3187764/view/3699044/app/users#T9bjQw9kLb7c

If you’d like to find those who have dropped off, you could switch “did” to “did not do”. Here is the cohort for reference: https://mixpanel.com/project/3187764/view/3699044/app/users#Umm8E4JZxLSe

I created two cohorts, Cohort AA: for users watched videos the last 30 days, and Cohort B: for users who liked videos in the last 30 days. I want to find out the users who did not watch videos AND did not like videos in the last 30 days. How do I create such a cohort?

You can do so by using the “AND” operator. Here is a cohort for reference: https://mixpanel.com/project/3187764/view/3699044/app/users#Tc76sqFRzbao

If you were to select multiple cohorts in the filter (example), we will filter by cohort A OR cohort B. Therefore, it is still possible for the following cases to occur:

User is not in Cohort A (fulfils the criteria), but is in Cohort B
User is not in Cohort B (fulfils the criteria), but is in Cohort A

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!