How users who made a purchase on your website behaved differently than the users who abandoned your shopping cart?

Both user segments had the intention to make a purchase on your website.

But one user segment chose to proceed with the checkout, the other choose to abandon your website.

What you can do to reduce the shopping cart abandonment?

How customers who came from the Valentines’ day campaign behaved differently than the customers who buy at other times of the year?

What was their average order value like?

Did they spend more money on the website?

If yes then how we can bring more users like them for other times of the year.

When is the best time to re-engage with your users? When is the best time for remarketing?

What is the rate at which you should acquire new users to maintain (if not increase) your website conversion rate?

Through cohort analysis you can get answers to such questions.

Welcome to the world of ‘User Disengagement’

We talk about user engagement all the time.

But we rarely discuss user disengagement.

User disengagement occurs when a user starts visiting your website less often, spend less time on your website, complete less goal conversions, generate little to no sales, unsubscribe from your newsletter and eventually fade into oblivion.

User disengagement is the harsh reality of your online business.

No matter what you do to retain your customers/subscribers/users, sooner or later they will move on.

If you send email newsletters to your subscribers, you know that you are going to loose some email subscribers with each new newsletter.

The bigger your subscribers base, more subscribers you are going to loose with each new newsletter.

I have seen businesses loosing hundreds of subscribers with each new newsletter.

I loose up to dozen subscribers and sometimes even more with each new email newsletter.

That used to hurt but after seeing the same pattern over and over again for so many businesses who do email marketing, I now have come to the conclusion that “user churn” is “normal”.

In the world of constantly changing technology (which give rise to better alternatives), emergence of new and powerful competitors and globalization it is just a matter of time and timing when you may end up loosing your customer / newsletter subscriber/ user etc for good.

Just look at MySpace.

It used to be top social networking website before Facebook came and took over.

While you can not completely stop user disengagement from occurring, you can certainly reduce the disengagement rate.

By focusing on user disengagement you can effectively reduce your customers/users’ attrition (also known as churn) and re-engage with them before it is too late.

You need to do lot more than re-engage with existing users.

You also need to constantly find new users.

While it is good to have returning users, you need lot more new users and that too all the time because majority of users eventually disengaged (visit your website less often, complete less conversions, generate less sales etc) from your website and become less profitable over time for no apparent reason.

While I loose up to dozen or more subscribers with each new newsletter, I also at the same time gain new subscribers which are 5 to 10 times more than the subscribers I lost.

So eventually I am gaining more subscribers with each new newsletter.

This should be the case with every business.

MySpace lost to Facebook because it failed to acquire new users.

It failed to innovate in a timely manner, so it also lost existing users.

So not only you need to constantly find new ways to re-engage with your existing users but you also need a constant flow of new users to compensate for the users lost over time and to maintain (if not increase) your website conversion rate.

You should not stop acquiring new traffic at any point and become content with the existing traffic, no matter how much traffic you are currently getting.

That would be like signing your own death warrant.

The Cohort analysis report in Google Analytics is all about user disengagement.

It is all about attrition.

It is all about the dreaded ‘churn’.

Understanding Cohorts

A cohort is a group/segment of users who showed common characteristics, attributes or experience in a particular time frame.

The characteristics/attributes of users are time bound because same users can show different characteristics in different time period.

For example a user may buy product X in Jan but buy product Y in Feb or a user may visit your website on Monday via Laptop but visit your website on Tuesday via smart phone.

Google Analytics define users’ characteristics/attributes in its reports through dimensions (primary and secondary dimensions) like traffic source, country, city, keyword, product, product category etc.

So all users who visited your website from a particular country belong to the same ‘country’ cohort.

All users who bought a product say product X belong to ‘product X’ cohort.

Similarly all users who visited your website for the first time on say feb 6, 2015 belong to the ‘feb 6,2015’ cohort.

A user can be a member of multiple cohorts at the same time depending upon how you segment and interpret the data.

For example

A user who visited your website from UK via Google Organic Search on February 14 and bought product X is a member of following cohorts:

• UK cohort – users of your website from UK
• Google traffic cohort – users who visited your website from Google
• Search traffic cohort – users who visited your website via search engine listing.
• Organic search traffic cohort – users who visited your website via organic search engine listing.
• February cohort – users who visited your website in February
• February 14 cohort – users who visited your website on February 14
• Product X cohort – users who bought product X

Cohort Analysis report in Google Analytics

This report (still in beta) is available under the Audience menu of your GA view:

You can analyze cohort’s behavior through any Google Analytics report.

But the one report which has been specially developed for analyzing cohort’s behavior of users is the cohort analysis report.

This report is esp. useful in understanding the behaviour of different cohorts in response to time sensitive/ short term marketing campaigns like Christmas sales, cyber monday, tv/radio ads, new email campaigns etc.

Understanding ‘Cohort Type’ in the Cohort Analysis Report

Cohort type is the dimension that characterizes the cohorts.

You can select only one cohort type at a time.

Use the Cohort Type menu to select a dimension:

At present only one cohort type is available in the Cohort Analysis ‘ report which is ‘Acquisition Date‘.

Consequently through cohort analysis report you can analyse the behaviour of only one type of cohort i.e. the group of users with same acquisition date.

Acquisition date is the date when users started their very first session on your website.

For example:

Day 0 => the day on which users started their very first session on your website. Also known as acquisition date.

Day 1 => the first day after the acquisition date.

Day 2 => the second day after the acquisition date.

Day 3 => the third day after the acquisition date.

Day 4 => the fourth day after the acquisition date.

Understanding ‘Cohort size’ in the Cohort Analysis Report

Cohort size is the size of the selected cohort.

The value of the cohort size depends upon the cohort type selected.

At present only one cohort type is available in the Cohort Analysis ‘ report which is ‘Acquisition Date’.

So cohort size has been defined only in terms of time frame (by day, by week, by month):

As Google rolls out more cohort types, the list of available cohort sizes is probably going to increase.

Cohort type: acquisition date ; Cohort size by day => all users who were acquired on the same day.

Cohort type: acquisition date ; Cohort size by week => all users who were acquired during the same week.

Cohort type: acquisition date ; Cohort size by Month => all users who were acquired during the same month.

Understanding ‘Metric’ in the Cohort Analysis Report

All the metrics in the cohort analysis report have been divided into 3 categories: ‘Per user’, ‘Retention’, ‘Total’:

Each category contains several metrics. Only ‘Retention’ category contains just one metric called the ‘User Retention’.

User retention is the percentage of users in a cohort who returned in the Nth day, week or month.

For example, percentage of users in a cohort who returned in the 2nd day or percentage of users in the cohort who returned in the 3rd day.

User Retention = number of users in a cohort who returned in the Nth day, week or month / total number of users in the cohort

Note: ‘User Retention’ is the default metric in Cohort Analysis report and you can only analyse one metric at a time in this report.

Understanding ‘Date Range’ in the Cohort Analysis Report

Date Range is the time period for which the cohort data should be displayed.

The value of the date range depends upon the ‘cohort size’ selected:

For example if the ‘cohort size’ is ‘by day’ then the value of ‘data range’ could be set to any one of the following:

1. Last 7 days
2. Last 14 days
3. Last 21 days
4. Last 30 days

Similarly, if the ‘cohort size’ is ‘by month’ then the value of the ‘date range’ could be set to any one of the following:

1. Last month
2. Last 2 months
3. Last 3 months

Interpreting the Cohort Chart

The cohort chart is a line chart which shows the cumulative metric values for the selected cohorts.

You can configure this chart through what Google called the ‘N selected menu’:

Through the menu you can select and compare up to 4 cohorts on the cohort chart:

You can apply up to 4 advanced segments (both default and custom) to the Cohort analysis report and these advanced segments are reflected in the Cohort Chart:

Interpreting the Cohort Data Table

Just below the cohort chart, you can see the cohort data table:

Each row in the cohort table represent a cohort.

‘Jan 31, 2015’ is the first cohort. It contains 4,845 users.

‘Feb 1, 2015’ is the second cohort. It contains 4,781 users.

‘Feb 2, 2015’ is the third cohort. It contains 8,546 users.

The date range you select decides the number of rows/cohorts in the data table.

For example, if you select ‘Last 7 days’ as the date range then the data table would contain 8 rows.

The first row or the top row shows the total or average value of all the cohorts for each column.

The remaining 7 rows show data for each cohort.

Similarly, if you select ‘Last 30 days’ as the date range then the data table would contain 31 rows.

The first/ top row shows the total or average value of all the cohorts for each column.

The remaining 30 rows show data for each cohort.

Each column in the Cohort Data table represents cohort size: one day/week/month of data.

The data table contains fix number of columns which is 13.

Each cell in the cohort data table contains the value of the metric you selected through ‘metric’ menu.

For example, if you selected pageviews metric then each cell would contain total number of pageviews per cohort per time increment.

If you selected ‘session duration’ metric then each cell would contain session duration per cohort per time increment:

The colour intensity in each cell visually indicates the magnitude of the metric value relative to other values in the cohort.

Cohort Analysis across segments

As mentioned earlier, you can apply up to 4 default/custom advanced segments to the cohort analysis report.

Each advanced segment appears as a separate data table in the cohort report.

If you have implemented enhanced ecommerce tracking on your website, you should apply enhanced ecommerce segments (like cart abandonment, no shopping activity etc) to your cohort report:

Here I have applied ‘cart abandonment’ ecommerce segment to the cohort report in order to understand when and where do these users disengaged and when and how I should re-market to them.

From the report above I can conclude that majority of users who abandoned the shopping cart do not engage with the website again.

Not even 1 day after the acquisition date.

So I have less than 24 hours to re-target them with new offer and increase my chances of getting the sales.

I also know from the report above that the users who came directly, engaged with my website a bit longer after abandoning their shopping cart.

You can get a similar insight through your cohort report and e-commerce segments.

Cohort Analysis and Data Sampling issues

I have noticed that the cohort analysis report suffer from data sampling issues much more than any other GA report esp. when I apply advanced segments to it.

I am not sure why is that.

It is a well known fact that the use of advanced segments create data sampling issues.

So if you run a high traffic website (10k sessions or more a day) and you do not use Google Analytics premium then avoid applying advanced segments to the cohort analysis report.

The advanced segments can greatly skew your cohort data. Instead look at the cohort analysis report in a filtered view.

If you have created the 10 Google Analytics Views that I have recommended repeatedly, then this should not be a problem.

Cohort analysis via User ID view

By default Google Analytics can not track users across devices/browsers as client ID can exist only on the device/browser where it has been set. In a user ID view you can get better users count.

Consequently your cohort analysis is going to be more accurate when done via the user id view.

To learn more about the user id view and cross device tracking, check out this article: Guide to Cross device tracking with User Id in Google Analytics

Advanced Cohort Analysis

You may have noticed it by now, that Google Analytics’ cohort analysis is still in its infancy and doesn’t provide the level of cohort analysis you may need to do.

If you want to do better and more cohort analysis then you also need to use other analytics tools like Kissmetric, RJ metrics, Tableau, ‘R’ etc which provide robust cohort analysis capabilities.

Other Articles on Enhanced Ecommerce Tracking

#1 Understanding Enhanced Ecommerce tracking in Universal Analytics

#2 Implementing Enhanced Ecommerce tracking in Universal Analytics – Nerd Guide

#3 Using Enhanced E-Commerce segments for Remarketing in Google Analytics

#4 Fixing Duplicate, Cancelled, Test orders & Refunds in Google Analytics

#5 Enhanced Ecommerce Tracking via Google Tag Manager