What is cohort analysis, and why is it important?
Let’s start by exploring what cohort analysis is NOT.
The main way people are used to analyzing data is by selecting a period of analysis from a calendar widget (eg. January 1-31), selecting some metrics (eg. visits) and dimensions (eg. goals) that need to be reported, and build a visualization for the metrics and dimensions that were performed during the selected period.
This method is the common method for analyzing specific events performed during a selected period.
However, when analyzing user behavior or conversion rates, the above method is often misleading and can lead to insights that generate more harm than good.
This is because accounts and users have ongoing behavior that will always be partially captured in the selected period for analysis when using the above method.
Let’s look at an example: it’s the end of the quarter and you need to report the performance of an advertising campaign that ran in the first month of the quarter.
You generate the report using the non-cohort method we described above, and you report the results:
When analyzing data the non-cohort way:
Conversion Rate: 30% (3 goals reached and 10 visits)
Important: 2 users that reached the goal didn’t visit on that day, offering a misleading image of what happened.
As you can see in the image, many more users converted outside of the selected period. Because of this, the campaign would be deemed as underperforming.
That’s why cohort reporting was introduced for analyzing user and account behavior, as it offers an image that mirrors reality when analyzing user behavior regarding any type of conversion or goal-reaching.
What IS cohort analysis?
A cohort refers to a group of users who experienced a common event in a selected time period.
A cohort analysis looks at the activity performed by that cohort of users even outside the selected period.
Here’s the same scenario, but analyzed using cohorts:
When analyzing data the cohort way:
Conversion Rate: 70% (out of the selected cohort of 10 users who performed a visit on a specific day, 7 of them reached the goal as follows: 1 on the same day, 6 reached the goal in the following days).
Cohorts are not only for retention
Most people are familiar with cohort retention reports that look like this:
Example of retention report - source: InnerTrends
The above visualization has the cohort of people that created an account in a specific week in the first column, and then reports week after week how many of them returned to the product. The higher the percentage, the darker the background color for that specific week.
Though very popular, the above is just one example of a cohort report visualization. In reality, almost every conversion report can be analyzed as a cohort report offering a much more accurate analysis.
Let’s see a practical example
InnerTrends is a cohort-based analytics platform. All the pre-built reports will let you / ask you to select a cohort: all of the accounts that share a common activity during a selected time period.
The conversions are then calculated automatically by looking at the activity that users/accounts performed at any time during or after the selected period.
For example, when trying to answer the question: “How many accounts were created and how many are now onboarded?” you will just need to select the period when these accounts were created. That’s the analyzed cohort.
InnerTrends will then look at the activity these accounts performed anytime during or after the selected period, and see which of them performed the event that marks the end of the onboarding process. All of them will be considered onboarded, regardless of when they finished the last step of the onboarding step.
Whenever you come back and look at this report, you’ll get an updated image of the onboarding conversion rate. That’s because over time, other accounts which were previously not yet onboarded, now may be onboarded. The only thing that will stay constant is the number of people that entered the cohort.
Because of this, cohort reports are the best way to analyze behavior and conversions, as they are the only reporting method that mirrors reality, and do not offer a distorted or misleading image.