A cohort is a group of users that share certain event together for a certain period of time (for example: users who make payment for the first time in the last 30 days, or added 7 friends in the last week). In other words, a cohort is a group of people with similar behavioral characteristics.
Cohorts let you group your users based on their actions, such as who performed add to cart event in the last 7 days but didn’t perform checkout event. In addition to ability of grouping users, you can see their progress over time.
List of users who belong in a cohort is automatically kept up-to-date, and you can segment funnels, retention, user profiles, flows, push notification and drill data based on cohorts you create.
Cohorts feature is available for Enterprise Edition customers.
Before diving deep let's get to know the plugin interface that consists of 3 main views: Overview, Detail and Compare.
Overview is the welcoming screen of Cohorts. As the name implies, it gives you some summarized information about your cohorts. In addition to that, you can create, edit or delete cohorts using this view. You can navigate to cohort detail by clicking its row as well.
- You can create cohorts using the button.
- You can navigate to Compare view.
- First column of the table: cohort name, last updated time and visibility setting of the cohort. You can add your cohort to favorites using star icon.
- 2nd column: Current number of users in your cohort
- 3rd column: The amount of change happened in the day.
- 4th column: 30-day trend of number of users in the cohort.
This view provides you an in-depth information about a cohort. While charts help you to understand enter/exit behavior of users, metric distributions show you a detailed breakdown of selected segments.
- You can change period.
- Depicts the number of users (net value) in your cohort for selected period.
- Depicts the entering and exiting users of your cohort for selected period.
- Shows metric distribution of your users.
- You can add/remove metrics using the button.
- You can edit the cohort using the menu.
Sometimes a side-by-side comparison of your cohorts might be handy. Compare view is designed to meet that need.
- You can select period.
- Here you select different cohorts to compare.
- Graph depicting total users in cohort graph
- Graph depicting total sessions of users in each cohort (data from drill)
- Graph depicting average session duration of users in each cohort (data from drill)
- Number of entering users for each cohort
- Number of exiting users for each cohort
- Table showing statistics for each cohort
Some notes on Cohorts before moving forward with how it works and how to add one:
- Cohort list, e.g which users are included in a particular cohort, is generated upon creation.
- This list is regenerated when you edit a cohort.
- The list is also regenerated at least once a day or at most once every hour, depending on the cohort generation time.
- Cohorts are not retroactive. Hence you will not see any users before a cohort is created, but only after you create a cohort.
A simple cohort
To create a cohort, go to Overview and define a cohort by clicking
Create cohort button.
Cohorts provide you a 2-fold segmentation mechanism. The first, user property segmentation, helps you to narrow your target group of users down using user properties. For instance, you may want only iOS users to be present in your next cohort. To do that, you can use the query builder in "USER PROPERTY SEGMENTATION" panel as demonstrated in below image. You can always extend your query to make your target set even more specific. If you don't want to specify any user properties, you can skip this stage too.
You can make your cohorts visible to everyone, or keep it private with optional sharing.
Once you are done with user properties, you can start adding behavioral conditions, using user behavior segmentation. First, you must select the people who have performed/not performed a certain events or sessions, and after that, you set the frequency and time range. When you complete entering these criteria, "add property" will become enabled, and lets you customize behavior condition.
You can add filters to many subjects such as country of a user, last entry time, input method, etc. and you can collect similar users. If you want to go deeper, you can increase the number of filters. The slight difference between this filtering and user property segmentation is that user property segmentation is not dependent on an event.
Using "+ Add Condition" button, you can add another conditions. As in the above example, you can observe users who log in using iOS and Facebook over the last 30 days and offer them special campaigns.
More complex behavioral scenarios
As you add more conditions, you'll see AND/OR selectors appearing between condition rows. For a user to be included in the cohort list, all of those conditions must be true. By selecting ORs, however, you can loosen up this hard requirement. If at least one of the items in the OR group is true, then others being false won't matter and final value of the group will be regarded as true.
Order of conditions doesn't matter.
This example reads as "Crashed at least once AND (didn't buy at least once OR didn't view at least once OR didn't start session at least once)". That is, only one of the conditions in the OR group being true is sufficient for that group to be considered as true.
Cohorts in action
As stated above, you can use cohorts to segment your users in user profiles, drill, retention, flows and many other plugins as well. Let's see some examples.
In the User Profiles section, you can observe the details of the cohorts more deeply. For example, you can get the most detailed information about the country distribution of users including Facebook login cohort, the platforms, devices, genders and ages of these users.Or you can observe users who have not logged in since March 25, you can reach them with a push notification and trigger them to use your app again.It is trivial to find different Cohorts + User profiles use cases. You can see the details of your customers in a single page who have purchased and shared products in your eCommerce application. Or you can find users who have added items to their cart, but haven't issued an order. It's all up to your business needs.
From user profiles page you can access the information about your users’ country, total time spent, last seen, as well as the device information used by your users. By observing the common features of your users with a particular pattern from a single page, you can make improvements in your product development process or in your content.You can also use a Drill segmentation to observe the retention rate of your cohort’s users. For example, you can observe the Retention information within 2 days of your registered users with Facebook Login in two steps as follows.Depending on the behavior of users, you can organize campaigns for these user groups by segmenting them, better understanding them, and you can make improvements to your product.
Another option is that you may want to use a drill to interact with your users based on the decrease or increase in engagement of your users. For example, your user group who logged in with Twitter has been using your application quite actively in recent days, and this number has fallen in the last week:
With a push notification you send your users, you can prevent this decrease and regain your users:
Or if you want to monitor your users’ User Flow, it is possible to do this again via cohorts. For example, you have defined a special Purchased & Engaged cohorts for your users who have bought a product from your app and have logged in again.
It’s possible to understand what steps your users follow and what they can not follow, and increase your sales by making improvements on your pages based on this information.
How can you benefit from Cohorts?
If you are a mobile app developer or product manager, you are very interested in all the actions users take with your mobile app. Specifically, you need to track the number of installs of a mobile application, the opening rates, clicks, whether the customer is interested, or uninstalled. You can use cohort analysis to evaluate the lifecycle of your mobile app.
In mobile related apps, a cohort is a group of users who complete a specific operation (installation, opening) within a certain period of time. You can evaluate a number of data samples, such as the proportion of active users, and we can also set up cohort events with various in-app events to determine more user interaction with your mobile application.
In e-commerce industry, cohort is a great tool to use. For example, to see the timeout period, revenue per visit, or average order value. Once you start doing cohort analysis you can look back over the past several months for specific groups of users.
These examples can be enhanced with gaming app, web app or social media app. In short, you will use cohorts as a tremendous tool in order to design the best possible benefits for your users.
What does Cohort Analysis provide for you?
Setting up a cohort analysis is a highly effective way of working since it helps you to distinguish customers separately. You can access the most accurate information by clustering and follow the individuals who are registered to your product during a certain period. In this way, the analysis and change over time of their interaction are not affected by the individuals in the other group; thus keeping the groups completely independent of each other.
“Engagement” vs “Growth”
Separating customers into cohorts is also effective in clearly identifying the difference between growth and engagement metrics. These two metrics can be confused with each other; because growth is the increase in the number of customers using a product or service. Generally, increasing numbers automatically increase general participation, but only new customers who access the product will probably stop after a while.
Comparison between different behaviors
A cohort analysis also helps in comparing the results between two or more groups. For example, if the “first seen on March” cohorts engaging the product more than the January cohort, an analysis of any changes that may occur between the two months may be necessary. In addition, more analysis can be done on the groups to see if the product is attracting a particular group.
A cohort analysis also helps to determine times when logins to the site decrease. Due to time-consuming work, quick decisions can be made to correct problem areas that may cause the decrease.