This document discusses behavioral cohorts and how they can be used to analyze user groups and monitor key metrics over time. It provides examples of common use cases like tracking user milestones, engagement with new features, conversion of new or non-active users. The document also demonstrates how to create different types of cohorts using Amplitude's tools, including inline cohorts with rolling windows and offsets to compare user behavior in different time periods.
9. In-line Rolling
Apply a Rolling Window to increase the size of the interval. Typically, a user would use a Rolling
Window to smooth out seasonality (accomplished by including more data points).
10. In-line Offset
Shift the time period of your cohort back in time using date inline offsets. Define transitions
between cohorts in order to understand where your growth is coming from!
11. Users who read 1 article in a week and then read 2+ articles the subsequent week. This
helps us target users who are finding value in our content and we can target them to
purchase a subscription.
Use Case: Media
12. How are current users (“any active event” in 2 consecutive periods) are engaging with
collaboration features in rolling 28 day periods.
Use Case: Productivity
13. Who are users that created an account last month and have become chart editors this
month? This group represents users that recently realized value in Amplitude, so it’s a
helpful metric for CSMs to watch to understand which accounts are growing healthily or if
trainings had an impact.
Use Cases: Amplitude
14. App: Users who installed the app, but did not make a purchase during the month
after install. This is important to track so that they can run re-engagement
campaigns.
Use Cases: Navigation App
15. Users who made a purchase in each day and also made a purchase at any point in
the prior 3 months. They view this metric as a better proxy for product satisfaction.
Previously, they’d have to use surveys to ask how likely users were to return to
make another purchase.
Use Case: Food delivery
16. “in each” cohort
+rolling by 90 days
1. Let’s create an “in each” in-line cohort for AmpliTunes users that made a purchase in March.
2. Let’s add a rolling 90 day window to increase the interval to include last 90 days (1/1 - 3/31).
So when we look at the 3/31 data point, it actually includes users who did Purchase 1/1 - 3/31 (90 data points
total because of 90 day rolling window).
Users who made a purchase in each day and also made a purchase at any point in the prior 90 days.
March
Purchase
1/1 - 3/31
FebruaryJanuaryDecember
Previous Period
17. “in each” cohort
+rolling by 90 days
+offset by 1 day
Users who made a purchase in each day and also made a purchase at any point in the prior 90 days.
March
Purchase
12/31 - 3/30
FebruaryJanuaryDecember
“Offset” is how you
can make something
happen “prior”
1. Let’s create an “in each” in-line cohort for AmpliTunes users that made a purchase in March.
2. Let’s add a rolling 90 day window to increase the interval to include last 90 days (1/1 - 3/31).
3. Add a 1 day offset to shift the date range into the past by 1 day (now 12/31 - 3/30)
Previous Period
18. “in each” cohort
March
Purchase
3/31
Let’s create an “in each” in-line cohort for AmpliTunes users that made a purchase in March.
If we take a single day, 3/31, users are counted if they made a purchase on that day.
Users who made a purchase in each day and also made a purchase at any point in the prior 90 days.
Current Period
19. “in each” cohort
+rolling by 90 days
+offset by 1 day
Users who made a purchase in each day and also made a purchase at any point in the prior 90 days.
March
Purchase
12/31 - 3/30
FebruaryJanuaryDecember
Purchase
3/31
4. Add the second Purchase event for our “current” period
Previous Period Current Period