In my last post I explained Latency Analysis, which is a canned analysis technique in Data Workbench. This post covers a lesser-known analysis technique called Cohort Analysis.

Cohort analysis entails segmenting a group of customers who share a common characteristic over a certain period of time. Cohort analysis becomes more informative when you can start to compare two or more groups over time.

Now that you have a basic idea of how cohorts work let’s walk through a basic cohort example so you can understand the value of looking at different customer groups over time.

In this example we will look at customers coming to your website from different referrers on November 1st, and the number of orders they have over time after that visit. This will help to identify the conversion effectiveness of different referrers, and how long a referrer needs for the visitor to convert. Below I will walk you through each step to show you how to create this analysis in Data Workbench, one of the capabilities within Adobe Analytics Premium.

number1  number2

Steps to Create a Cohort Analysis

Step 1: First, create a customer segment.  Based on the example I gave above we will look at the different referrers that came to your site on November 1st.  So we will open a Visitor level segment in Data Workbench.


Step 2: To ensure that the Visitor has both the conditions of coming from a specific referrer on a certain day, you need to first create a visit level segment, with the selections in the referrer table and the time table (day/week/month) and then create a Visitor level segment based on it.  Remember, the Visitor segment dimension is exclusive and the order in which you create it matters.  So if we create a Google segment first, the customers in that segment can’t be in any other subsequent segment, when using the Visitor segment visualization. (There are other approaches if needed for analysis, but that would be the subject of another post.)


After you create a segment, remember to rename it so you know which visitors are in that segment.  You will need to repeat this step for all of the segments that you want to create.

Step 3: Once you have created all three visitor segments you will save the segment as a dimension.


The dimension will be saved in your Data Workbench user folder.  Once the dimension is saved you will be able to find that dimension in table menu structure and use it in various other visualizations and analysis techniques within Data Workbench.

Step 4: Now that the dimension is created, open a graph visualization based on the time segment that you selected above.  For our example, select the day graph visualization.


Next, change the display from bars to lines so you can see the metrics over time.


Step 5: Once the graph is changed to be a line graph, change the series in the graph to the dimension you created for the referrer segment.



Open the dimension outside this graph so you will be able to understand which line represents each segment.


Step 6: Now we are able to see what these specific groups are doing, over time, after the time and event (referrer selection) criteria.  Since the Google segment is much larger than the other two in this scenario zoom into the Yahoo and MSN segments.  To zoom in on a graph, hold down the left and right buttons on your mouse and drag your mouse up or down.  Now that we are zoomed in on Yahoo and MSN, add the metric of orders.  To do this, right click on the Visits metric and select “add metrics”, and then select “orders”.  Here is what the graph will look like after you have done the above selections.


You can see in the above graph there is a spike (on the day you selected) when creating this visitor segment.  The days following the spike are what those customers from each segment are doing.  You can see that there are different spikes throughout the 30 days after November 1st.

When we dig deeper into this example we see that the MSN visitor segment has a bump about 4 days after November 1st, for orders, then doesn’t really spike until about 30 days later.  The Yahoo customers have a couple of small lifts 6 and 12 days after November 1st, but they have a larger lift days later (Nov. 19th).


This is the power of cohort analysis: you are now able to see which referrer is best at converting customers from a certain point in time. If your website changes on a regular basis, using cohort is a must because one change to your site can impact different customer segments greatly. This also allows you to compare different KPIs based on when the website changed. Selecting the date the site changed and the user group you want to analyze.

As evidenced from the above steps, this is a more complex and time consuming analysis technique than Latency analysis, but it is able to provide some additional analysis points.  Once you get the hang of doing Cohort analysis you will find it to be another great tool to add to your Data Workbench tool belt, and will be able to apply it to many different dimensions of analysis.  If you have any questions or comments on how to do a certain analysis please feel free to leave them below or reach out to me on Twitter at @adaste.