Getting Started with Cohorts

The Cohort tool allows you to understand how often your customers return and engage with your website or product.

• Analyze drivers of user retention as well as the factors related to churn
• Create user cohorts based on repeated behaviors and attributes
• Identify customers with high lifetime value (LTV)

Build your Cohort Analysis

Define your Cohort

To begin a cohort query, determine an initiating event (called the Cohort event). The first event of a cohort is required; a user must complete the initiating event or Cohort event and then return to perform a second event which is explained in the section “Target Behavior”. Custom Events and Merged Events can be used. As with other tools, you may apply a Filter Where.

You may select a different time zone from your project time zone on a per query basis by locating the globe icon on the top right of the query screen.

You can chain multiple events in a sequence using an “and then performed” clause to define your cohort of users. In the following example, PetBox wants to measure users who download their app, and then start the app.

Generations and Breakouts

In addition to the first event, cohorts can be defined by a shared generation or a shared property. A generation is a unit of time, such as a month. A monthly cohort would include all users who entered the cohort during that month. A property is a characteristic or attribute, such as device type. Cohorts defined by device type would include all users with an iPhone, all users with an Android, etc. A user will only appear once in the results of a cohort analysis. For generation cohorts, users will be put into the property breakout in which they first appear during the time interval.

Target Behavior

After selecting an initiating event, you must select a Target Behavior event. This second event of a cohort is also required; Custom Events and Merged Events can be used. As with the initiating event, you are also able to apply a Filter Where.

• Event: Often this is an event that is repeated multiple times, such as a purchase. This is the event that represents the subsequent user behavior that you wish to analyze.

• Revenue: Target behavior can also be represented as revenue. Using revenue as the target behavior will analyze the revenue generated over time by each cohort.

  

Date and Time Range Settings

Every query requires you to select a date range. In Cohort analysis, the date range refers to the time period during which a user completes all steps of the cohort query, defined in Row A. All new queries default to Last 30 Days. To open the date range selector dropdown, click on Last 30 Days. The start date is the first day to be included in the search. The end date is the last day. As mentioned in Cohort Basics, you can set the time zone for your Cohort query.

Generation

Every cohort is defined by a first event, a breakout, and a second event. To appear in the results, a user must complete the first event and also have a defined value in the breakout. Breakouts may be an Event Property, a User Property, or a User Segment; or they may be a Generation; or they may be a combination. A generation is a time-based grouping that describes the cohort. It describes when the first event occurred. You may select Hour, Day, Week, or Month. For example, to create monthly cohorts of users using the date they signed for a newsletter, the first event would be Newsletter Signup and the generation would be Monthly. This will produce a list of all users who signed up for the newsletter, broken out by month: January signups, February signups, March signups, etc.

Date Range

To select fixed start and end dates, use the date range selector on the left side of the dropdown, you may select a start date and end date from the calendar. You can also enter a specific date by selecting on the date at the top of the selector and entering a value. Use the left and right arrows to navigate the calendar. Tick the Today checkbox to create a dynamic end date.

The right side of the dropdown lists all of the dynamic date ranges that are available. You may choose a dynamic date range, for example Last 7 Days or Last Full Month. This will automatically update the date range of your query each time you view it, counting backwards from today. If you select Last Full Week, then Analytics will analyze the most recent complete week, defined as Monday to Sunday. If you select Last Full Month, then Analytics will analyze the most recent complete month. You can quickly navigate the calendar to select full months using the links in the lower left corner of the dropdown.

To save a custom date range, for example Last 45 Days, simply click Add Custom Date Range in the lower right corner of the dropdown. Your previously used custom date ranges will be saved for future use and are viewable alongside the default dynamic date ranges.

Interval

In Cohort, the interval describes the time period during which a user completes the Target Behavior, defined in Row C. You may choose hourly, daily, weekly, or monthly intervals. Interval options are dependent on the generation and date range selected.

For example, if the selected date range is Last 7 Days, then the available time intervals will be hourly and daily. Weekly will not be available because there is only one week in a seven day date range. If the date range is changed to Last 90 Days, then the intervals dropdown will update with the additional options of weekly and monthly, but will no longer allow hourly analysis due to the inability to visualize.

To calculate the total event count or user count across an entire date range, match the duration and interval of the report. Hourly intervals are only available if a cohort is defined by an hourly or daily generation.

In the results section, intervals are listed as columns. Interval 1 is the same interval during which the first event was completed. Interval 2 is the next interval after, and so on. If the interval is set to daily and a user completed the target behavior within the same day, then they will appear in Day 0.

Note: For Cohort queries with the time interval set to Daily, “Day 0” captures users who completed the target behavior in the subsequent 24 hours after entering the cohort (meeting the definition for inclusion).

For example, a user who performed Site Visit on Jan 1 at 1pm (and thus entered the cohort) would be included in Day 0 as long as they performed Subscribe (the target behavior) at any time before 1pm on Jan 2. Events are tracked down to the millisecond, and will be categorized in intervals by this level of precision.

Recurring vs First-Time

You may adjust your cohort query to observe the frequency of recurring target behavior or to measure the first time that the target behavior occurred after completing the first event. If a query is set to Recurring, then a user will appear multiple times within a row if the user repeated the target Behavior multiple times. If a query is set to First-Time, then a user will appear only once in a row, describing when they completed the target behavior.

To exemplify the difference between recurring and first-time cohort queries, consider a box subscription company that wants to measure customer behavior among their newsletter subscribers. The first event could be “Newsletter Sign Up” and the target behavior could be “Purchase”. User A signed up for the newsletter in January, and then completed purchases in March, April, and May. In a recurring cohort query, the user will appear in the January cohort row, and they will appear in the Month 3, Month 4, and Month 5 columns. In a first-time cohort query, the user will appear only in the Month 3 column. Thus, exclusivity of cohorts as they appear in the interval counts only occurs in a first-time cohort. Recurring cohorts can have a user appear in multiple intervals, though the total number of users will be exclusive.

Cumulative vs Non-Cumulative

All new queries default to the Non-Cumulative setting. Non-cumulative cohort queries show the count or percentage of users in the cohort who performed the target behavior within the interval. Cumulative cohort queries show the count or percentage of users in the cohort who performed the target behavior as a running total over time. Cumulative counts are only available for queries measuring first-time behavior. For a full explanation of cumulative and non-cumulative, see Cumulative vs. Non-Cumulative Analysis in Cohort.

• Non-Cumulative Percent: Displays the percentage of users who completed the target behavior for each selected time interval or breakout.
• Non-Cumulative Count: Displays the count of users who completed the target behavior at each point in time defined by the selected time interval.
• Cumulative Percent: Displays the percentage of users who completed the target behavior for the first time as a running total over time.
• Cumulative Count: Displays the count of users who completed the target behavior for the first time as a running total over time.

Visualization Options

Cohort analyses have four different visualization options: Circle Heatmap, Heatmap, Line Chart and Area Chart. You can toggle between these options in the visualization dropdowns. You can also download a cohort analysis as a CSV file.

Annotations

Cohort annotations act as general notes about the cohort analysis over the entire designated date range. To add an annotation to Cohort, click on the Annotation flag icon in the query builder window and click Add an Annotation. To access existing annotations, click the Annotation icon in the Data Panel. For more information about annotations, visit this article.