Most companies aren’t going deep enough with their user data.
Conversion funnels, retention, and segmenting by install cohort help you identify where your users are getting stuck or leaving, but aren’t very helpful in telling you what you should do instead.
In this session from Opticon 2015, Spenser Skates, CEO and Co-founder of Amplitude, discusses how you can use behavior-based cohorting to maximize user engagement and retention. By grouping your users based on what actions they take, you'll discover how you can improve your product to drive growth.
As with Facebook’s famous ‘7 friends in 10 days’, learn how you can find the ‘magic moment’ to focus your product around.
Watch the video here: https://amplitude.wistia.com/medias/hdpnipul1l
Actionable Cohort Analysis: How to use behavior-based cohorting to maximize engagement and retention
1. Know your users. Grow your app.
Actionable Cohort Analysis:
How to use behavior-based cohorting to maximize engagement and retention.
Spenser Skates
CEO and Cofounder, Amplitude Analytics
3. Level 3: What drives growth
Level 2: Where are my problems
Level 1: Counters
Behavior-Based Cohorting
Data Science
Companies at this level:
Facebook, Amazon, Netflix, Zynga
Funnels
Retention
Segmentation
Events
DAU, MAU, Revenue
The Three Levels of Analytics
4. Level 3: What drives growth
Level 2: Where are my problems
Level 1: Counters
Behavior-Based Cohorting
Data Science
Companies at this level:
Facebook, Amazon, Netflix, Zynga
Funnels
Retention
Segmentation
Events
DAU, MAU, Revenue
The Three Levels of Analytics
5. Level 3: What drives growth
Level 2: Where are my problems
Level 1: Counters
Behavior-Based Cohorting
Data Science
Companies at this level:
Facebook, Amazon, Netflix, Zynga
Funnels
Retention
Segmentation
Events
DAU, MAU, Revenue
Most companies are here.
The Three Levels of Analytics
6. Completing 50% of Profile
Adding 7 Friends
What Drives Long Term Retention?
Posting on a wall
Optimized experience around adding and discovering friends.
7. Properties that have a review
Properties with
Professional Photography
What Drives Revenue?
Properties with a
personalized description
Shifted from a search based to a browse based
experience to encourage aspirational travel
8. Level 3: What drives growth
Level 2: Where are my problems
Level 1: Counters
Behavior-Based Cohorting
Data Science
Companies at this level:
Facebook, Amazon, Netflix, Zynga
Funnels
Retention
Segmentation
Events
DAU, MAU, Revenue
Most companies are here.
The Three Levels of Analytics
9. Level 3: What drives growth
Level 2: Where are my problems
Level 1: Counters
Behavior-Based Cohorting
Data Science
Companies at this level:
Facebook, Amazon, Netflix, Zynga
Funnels
Retention
Segmentation
Events
DAU, MAU, Revenue
Most companies are here.
The Three Levels of Analytics
10. Level 3: What drives growth
Level 2: Where are my problems
Level 1: Counters
Behavior-Based Cohorting
Data Science
Companies at this level:
Facebook, Amazon, Netflix, Zynga
Funnels
Retention
Segmentation
Events
DAU, MAU, Revenue
Most companies are here.
The Three Levels of Analytics
12. What is a behavioral cohort?
In traditional analytics, a “cohort” has referred to an
Acquisition Date Cohort:
A group of users who started in the same time period, and are then
tracked together over time.
13. What is a behavioral cohort?
Acquisition Date Cohorts can tell you that users who joined in June are
retaining better than users who joined in May…but it’s hard to take action on
that information.
14. What is a behavioral cohort?
Behavioral cohorting allows you to group your users based on specific actions
that they have or have not taken in your app or website.
This helps you identify user actions or product features in your app that drive growth.
15. Compare retention between behavioral cohorts
Users who add 7 Friends in 10 Days. Users who do NOT add 7 friends in 10 days.
100%
80%
60%
40%
20%
0%
Day 14
Users who add 7 friends in 10 days: 36.1%
Day 0 Day 1 Day 3 Day 7 Day 14 Day 30
Product Decision: Optimize Facebook experience around adding and discovering friends
16. How to use behavioral cohorting to make product decisions
Come up with a
hypothesis on what
user actions lead to
long term retention
Apply cohorts to
retention reports
and search for
insights
Test the hypothesis
by creating cohorts
based on specific
user actions
Step 1 Step 3Step 2
18. Step 1: Come up with a hypothesis
Example: Social music app
Key user actions on the 1st day:
• Joining a community
• Completing user profile
Come up with a
hypothesis on what
user actions lead to
long term retention
19. Step 2: Create cohorts to test your hypothesis
Cohort: Users who joined at least 1 community on the 1st day of use
20. Step 2: Create cohorts to test your hypothesis
Inverse Cohort: Users who did NOT join at least 1 community on the 1st day of use
25. Step 3: Apply cohorts to retention reports
Takeaway: Users who join at least 1 community on the 1st day of app use have
significantly increased 7 and 14 day retention compared to users who do not join a community
26. Now we can repeat this process for our other candidate action:
Completing the user profile on the 1st day
28. Takeaway: Users complete their profile on the 1st day of app use have
significantly increased 7 and 14 day retention compared to users who do not complete their profile
Cohort: Users who completed their profile on the 1st day of use
29. Both joining a community and completing a profile on the
first day correlate with improved retention.
Let’s compare the 2 pairs of cohorts on the same graph.
Now what?
30. Takeaway: Joining a community has a greater effect on retention than completing a profile.
31. Takeaway: Joining a community has a greater effect on retention than completing a profile.
Product Decision: Encourage users to join a community during onboarding experience.
32. • Unlike traditional acquisition date cohorts, behavioral
cohorts group users by specific actions they take in your
product
• Behavioral cohorting allows you to discover the actions
that can maximize long-term retention, just like Facebook’s
“7 friends in 10 days.”
• There are 3 steps to behavioral cohort analysis:
• Come up with a hypothesis of actions that promote retention
• Create cohorts to test that hypothesis
• Apply the cohorts to retention or engagement reports
Summing Up
‘Analytics’ at this point is such a noisy term – let’s break down what we mean when we talk about analytics.
‘Analytics’ at this point is such a noisy term – let’s break down what we mean when we talk about analytics.
First define traditional cohort, then what a behavioral cohort is.
First define traditional cohort, then what a behavioral cohort is.
First define traditional cohort, then what a behavioral cohort is.
Retention graph
First define traditional cohort, then what a behavioral cohort is.
First define traditional cohort, then what a behavioral cohort is.
First define traditional cohort, then what a behavioral cohort is.
This is the retention module – daily (Nth day) retention for new users over the last 30 days. We can add our cohorts in the USER SEGMENTS sidebar to compare retention between cohorts.
This is the retention module – daily (Nth day) retention for new users over the last 30 days. We can add our cohorts in the USER SEGMENTS sidebar to compare retention between cohorts.
This is the retention module – daily (Nth day) retention for new users over the last 30 days. We can add our cohorts in the USER SEGMENTS sidebar to compare retention between cohorts.
This is the retention module – daily (Nth day) retention for new users over the last 30 days. We can add our cohorts in the USER SEGMENTS sidebar to compare retention between cohorts.
This is the retention module – daily (Nth day) retention for new users over the last 30 days. We can add our cohorts in the USER SEGMENTS sidebar to compare retention between cohorts.
First define traditional cohort, then what a behavioral cohort is.
First define traditional cohort, then what a behavioral cohort is.
This is the retention module – daily (Nth day) retention for new users over the last 30 days. We can add our cohorts in the USER SEGMENTS sidebar to compare retention between cohorts.
Feel free to reach out w/ questions
Both of the inverse cohorts (not joined community and not completed profile) have similar retentions.
Both of the inverse cohorts (not joined community and not completed profile) have similar retentions.