Our cohort report shows you the people who progress from doing one event to doing another event. It also shows you if people are repeatedly doing a single event.
What sets it apart from a regular conversion funnel or metric is that you can further segment these people by when they did the first event, or segment them using your typical Kissmetrics properties (the method for segmenting any other report). These different segments are called cohorts, hence the name of the report. Then you can see how your cohorts perform the second event over a long period of time.
Every Kissmetrics account has access to a basic Retention Report, which tells you about visitors who repeatedly come back to your site.
Advanced cohort analysis and the ability to create multiple cohort reports is generally reserved for paying customers only.
The easiest way to show how to read the cohort report is through an example. Let’s look at an example of a Default Retention Report. Using the Retention Report in every account, let’s analyze the people who:
- Visited Site (event 1)…
- From October 19, 2019 - March 11, 2020 (date range), and…
- Group these people by week (grouping).
- Of these people, analyze how many of them Visited Site again (event 2)…
- For each subsequent occurrence of Visited Site (count), and…
- Group/bucket these people by how many weeks it took them to do Visited Site again (buckets)
To say this again in one sentence, we’re looking at the people who Visited the Site from October 19, 2019 to March 11, 2020 (organized by which week they Visited Site), and of these people, seeing what percentage of them returned to Visit Site again, week over week.
Running this report gives you something like the graph below. Let’s look at this one piece at a time
Let’s start with how the cohorts are created. This is the information in the leftmost columns.
Three options go into creating this area:
- Event #1: Visited Site
- Date Range: October 19 to March 11, 2020
- Grouping: group across weeks
Essentially, we do a people search for all the people who Visited Site from June 1 to March 11, 2020. Then we create the groups like this:
- Of all of these people, take the people who Visited Site during the first group of the date range (Week of Jan 1). Assign these 35 people to Cohort #1 for the Week of Oct 19.
- Of the remaining people not in Cohort #1, take the people who Visited Site during the second group of the date range, the week of Oct 26. Assign these 6,790 people to Cohort #2 for the Week of Oct 26.
- Of the remaining people not in Cohorts #1 or #2, take the people who Visited Site during the third group of the date range, the week of Nov 2. Assign these 11,949 people to Cohort #3 for the Week of Nov 2.
- And so on, until you’ve finished grouping the entire date range.
Now to the meat of the report.
Three options go into creating this area:
- Event #2: Visited Site again
- Count: count every time they Visited Site again after the first instance
- Buckets: divide people into weekly buckets
For each cohort, what percentage of people Visited Site again within 1 week of doing the first event? Within 2 weeks? 3-8 weeks?
- For Cohort #2 (the second row), 30.4% of people came back to perform Visited Site again within 1 week (0-6 days).
- For Cohort #2, 18.3% of people came back to perform Visited Site again within 2 weeks (7-13 days)…and so on.
- For Cohort #3 (the third row), 21.7% of people came back to perform Visited Site again within 1 week (0-6 days).
- For Cohort #3, 10.3% of people came back to perform Visited Site again within 2 weeks (7-13 days)…and so on.
In fact, you are able to click into each cell and get a list of say, the people from Cohort #2 who returned within 3 weeks.
When looking at a cohort report, it’s clear what the date ranges for the cohorts are, but not the buckets. In the above example, Cohort #1 is well-defined as being people who Visited Site between Oct 19-25, 2019. Consider the bucket for repeat Visits between 7-14 days of the first visit - Bucket #2. You might expect this bucket to nicely cover the week of Oct 26, 2019 (Oct 26-Nov 1), but Bucket #2 actually includes more days.
Let’s say Bob originally Visited Site on October 24. He is placed in Cohort #1. 7-14 days after this particular visit would actually span Oct 31- Nov 7. If Bob Visited Site again in this time frame, he would show up in Bucket #2. Because of this behavior, Bucket #2 (repeat Visits between 7-14 days after Cohort #1’s first visit) would thus actually span Oct 26-Nov 8.
This guide explained how to read a cohort report, but these articles will give you better ideas for how to use a cohort report:
Joshua Porter - “Cohort Analysis - Measuring Engagement Over Time”.
Dan Martell - “Cohort Metrics for Startups Revealed – Part I: Plain English”.
Dan Martell - “Cohort Metrics for Startups Revealed – Part Ii: Aged Groups”.
Andrew Chen - “How to measure if users love your product using cohorts and revisit rates”.
Updated about a year ago