Behavior offset: Segment users over two distinct time periods

With Amplitude's behavioral cohorts, you can create groups of users who share a behavior pattern. The behavior offset feature segments these users based on behaviors they displayed in two distinct time periods.

Common use cases for behavior offsets include:

• Identifying users who made at least one purchase during the past week and at least two more purchases over the preceding 90 days. This is a useful proxy for customer satisfaction.
• Measuring users who installed the app but didn't make a purchase during the 45 days after install. This information can help you target re-engagement campaigns and convert new users into paying customers.
• Identifying users who purchase a subscription but don't read an article in the following two weeks. Media companies can use this segment for churn prevention efforts.
• Identifying users who read one article in a given week and then read two or more articles the following week. You can target these engaged readers for subscription offers.

In each of these use cases, there are two cohorts in play, separated from each other by the passage of time. In the second example, they are:

• Users who installed the app, and
• Users who didn't make a purchase in the subsequent 45 days

Behavior offsets segment the users who appear in both cohorts.

Before you begin

Before you get started, review behavioral cohorts and rolling windows.

Add a behavior offset to an in-line cohort

This section follows the first example in the bulleted list above: segmenting for users who made at least one purchase in the current period, but also made two or more purchases in the last 90 days.

To add this behavior offset, follow these steps:

1. In the Segmentation Module, click + Performed to begin creating a new in-line cohort.
2. Define the previous period event. In this example, the previous period event is Complete Purchase.
3. Adjust the operator and event frequency, if necessary.
4. From the More Options menu for this event, click Add rolling.

In the rolling over field, enter the number of days you want to allow for the current period event to occur. In this example scenario, that value is 90.

When using a during cohort instead of in each, the offset function lets you create daily offsets, regardless of the interval you set in the date picker. Refer to in-line cohorts in Amplitude for more detail.

5. From the same menu as in step 4, click Add offset. In the offset field, enter the number of days after the occurrence of the previous period event that you want to wait before the rolling window (see step 4 above) begins.

   In this example, the day of the previous period event doesn't belong in the rolling window, so enter 1. If you want to include that day in your analysis, don't include an offset.

6. Click + Performed again. From the and who performed drop-down, select the current period event. Continuing with the example, this is Complete Purchase.

Applying the rolling window and offset to the previous period event shifts the cohort to an earlier time period. This process offsets only the in-line cohort, not the event selected in the Event Module.