Home | News | Cohort Analysis: Its Full Potential

Back to the Board

Data & Analytics

  • Posted on 19 May 2016

Cohort Analysis: Its Full Potential

The Cohort Analysis report in Google Analytics, has been out for a year now, so it’s about time to see the full potential of this report!

What is a cohort?

A cohort, is a group of users that share a particular characteristic during a certain time span. At first glance a cohort looks pretty much like a segment. The biggest difference being, the time dimension of the cohort. Although a segment and a cohort seem to have similar functions, the best way to use the Cohort Analysis report, is by using them together. In this article, we will point out some useful applications of segments within the Cohort Analysis report.

Within the Cohort Analysis, there are a couple of options to select from. Since the Cohort Analysis report is still in beta, some options are however limited. For instance, the only ‘cohort type’ that can currently be selected, is ‘acquisition date’.




Aside from the type of the cohort, you can choose its’ size by selecting day, week or month. This selection relates to the last option, where you can define the date range of the analysis. Cohort size ‘by day’ has a range of seven to thirty days, ‘by week’ a range of one to twelve weeks, and ‘by month’ a range of one to three months. With the third option, you can choose the metric that will be applied to the cohort. This can be revenue, page views, sessions, goal completions, but also user retention which is a great metrics to analyse churn.

When you look at the graph when applying the user retention metric, the trend seems a bit odd at first. It makes a steep drop from 100% on day 0 (which is the day of acquisition), to just a couple of percent on the next day. The downfall is actually quite normal, because it shows how many visitors came back to your site on the days following acquisition day. The graph shows that three percent of the visitors came back after one day. It’s more difficult to see the differences between the following days because returning users are marginal compared to acquisition day. The tables below the graph are therefore more useful.




With the data from the tables, you can adjust your remarketing strategy. You can for instance, decide to start remarketing on the day of the biggest decline. Another conclusion that could be drawn from the data, is the length of the decision making process.

The most interesting way to use the Cohort Analysis report, is adding segments and comparing these to the ‘all sessions’ data. You should look for segmented cohorts that show significant differences to the ‘all sessions’ data. Any difference, positive or negative, can lead to an insight into time-based group of visitors that you can optimise on. 


The tables below show a comparison between mobile/tablet traffic and desktop traffic. By setting the date range to three months and selecting revenue as the metric, you can see the overall differences between the devices. It now becomes easy to see that many users purchased something in the first month after they visited the website. A comparison between the device types also shows that this behaviour is similar on all device types. 



An insightful way of applying segmentation in the Cohort Analysis report is by creating a funnel based on the metric user retention of ‘all sessions’, ‘CPC traffic’ and traffic from a specific campaign. The graph below shows that CPC is pretty much the same as ‘all sessions’, but the percentage of users returning in the next day’s coming through Campaign 1 is much higher than the average of ‘all sessions’. This is a campaign that was live for just a couple days. That is why last 7 days are selected as a date range. Should you want to analyse long term CPC campaigns then it’s better to select the last 3 months as a date range. 




Traffic Sources

The most useful segments are often segments that represent the different traffic sources like direct, organic or email. Each segment will give new insights into the effectiveness of that particular traffic source.

A segmentation on email traffic in the ‘Acquisition reports’ shows the total amount of visitors. But within the Cohort Analysis report it will show when these visitors come back to the website. This will give you a whole new way to track the success of email campaigns.



To summarize, the Cohort Analysis report can give you a lot of data and new insights. Especially when you use it in combination with standard and custom segments. Note that this will probably produce some sampling in your reports, but that shouldn’t affect its usefulness.  Cohort analysis is all about detecting irregularities and thus opportunities for optimization, and not so much about reporting exact numbers.

The Cohort Analysis report in Google Analytics is currently still in beta. It will probably soon be extended with new segmenting options, especially regarding the cohort type. We for one are looking forward to creating cohorts based on first or most recent purchase date or other business specific conversions, like signing up for a membership or downloading a white paper. That will give us a true insight in how clients that share the same time-based characteristics interact with our website.


If you want to discuss your digital analytics strategy for 2016, get in touch 

NetBooster Q1 2016

Sven Buning

Data & Analytics Consultant, Netherlands




Posted by NetBooster (Group)