Cohorts

Intro to cohort analysis

Cohort analysis is a method to track how different groups of users behave over time. A cohort represents a group of users with shared characteristics, for instance:

• Users who began using your product in a specific month
• Users from a particular country or segment

While a cohort can also refer to (for example) a group of products, it's usually about users.

Cohort analysis involves examining how these groups of users behave over time. For example, you might observe how many of them return each month, or how much they spend over time.

Some scenarios where you might use cohort analysis:

• Activation analysis - Group users by their start date and monitor their activation progress over time.
• Retention analysis - Group users by their start or activation date and observe how many continue to use the product over time.
• Revenue analysis and customer lifetime value - Group users by their start or first purchase date to identify trends in revenue and customer lifetime value.

Exploring cohorts

How to read a cohorts table

• Row by row - Each row gives you insights into the behavior of user groups over time. For instance, you can see how long different user groups remain active and at what point you start to lose them.
• Comparing rows - Comparing rows gives insights about your product's lifetime. You can compare the behavior of user groups who began using your product early with those who started recently. This can give you insights on things like product updates or marketing efforts.
• Diagonals - Any diagonal pattern in your cohorts table can provide insights into seasonality or indicate that something significant happened at a specific time. This is because calendar periods, for example 'May', will always be represented diagonally in a cohorts table.

Breakdown

Choose how to break down your cohorts, grouping user cohorts by a start date or a dimension. You can also adjust the granularity of both rows and columns.

Filters

Your cohorts table can be filtered like any other visualization.

Defining cohorts

To add cohorts to a metric, use the Cohorts section in the metric definition. As a workspace administrator, navigate to the metric and select 'Edit'. Once defined, cohorts are visualized as a table.

There are two ways of defining cohorts in Steep:

• Standard cohorts – Easily define cohorts from your existing tables, dynamically calculated for quick insights. No adjustments to your data structure required.
• Advanced cohorts – Built from tables explicitly created for cohorts, supporting more complex analyses like advanced retention cohorts.

Standard cohorts

Standard cohorts are defined from general detailed tables, for example orders or transactions, and dynamically calculated. They are flexible and provide a great option for most basic use cases.

Setting up a standard cohort

1. Define your metric basics and add Cohorts using the + button
2. Choose a Cohort time - this defines how cohorts will be grouped
3. Label your cohorts
4. Decide if you want to break down by calendar period or rolling period (e.g., 7, 30, 90, or 365 days).
5. Let the % of initial value be toggled off (this is used for standard retention cohorts)

Setting up a standard retention cohort

1. Set it up like a standard cohort. Use a join if necessary.
2. Toggle on % of initial value to create a breakdown where the current value for each column is divided by the cohort's initial value.

Dimensions

Any defined dimension can be used to group cohorts.

Rolling period or calendar period

By default your cohorts table will breakdown your cohort over calendar periods, but you can choose to change this to rolling periods (7, 30, 90 or 365 days)

Granularity

For standard cohorts, all workspace members can automatically explore all granularities of the cohort table, for both rows and columns. If needed, it is possible to limit the granularity options, for the cohort table, in the metric definition.

Adjusting the granularity will impact both the rows and columns.

Cohort only

Enable the Cohorts Only option for metrics that are only used for cohort analysis. An example of this could be a user retention metric. This option disables all other visualization types (line, bar, rank and pivot table) for the metric.

Advanced cohorts

Advanced cohorts are built from a pre-aggregated table created specifically for the cohorts you want to set up. They offer less flexibility but support more advanced use cases, such as more advanced retention cohorts.

Granularity

Decide on one specific granularity when pre-aggregating a cohorts table, for example daily or monthly. Use the Granularity option to set Steep cohort definition to the same granularity. Adjusting the granularity will impact both the rows and columns of your cohorts table.

Setting up advanced retention cohorts

1. Prepare a new cohorts table by combining data from other tables and pre-calculating cohorts values. You have full control over how to calculate your numbers; for example, you can make them cumulative or remove deleted accounts from the relevant periods, among other options.
2. Define your metric in Steep by referencing columns in your prepared cohorts table.
3. Define what granularity your prepared data is stored in.

Cohort only

Enable the Cohorts Only option for metrics that are only used for cohort analysis. An example of this could be a user retention metric. This option disables all other visualization types (line, bar, rank and pivot table) for the metric.

Data source support

Supported

Cohorts functionality is currently supported using any of the following data sources:

• Athena
• BigQuery
• Databricks
• PostgreSQL
• Redshift
• Snowflake

Not currently supported

The following data sources are currently not supported for defining cohorts:

• MySQL
• Microsoft SQL Server
• Synapse SQL