Kin + Carta is joining Valtech. Learn more here >

Skip to main content

Select your location

Search
Black background, abstract web illustration of colours green, pink, red

Supercharge your Google Analytics with Machine Learning

Understanding the people who visit your platform is more crucial than ever.

With insight into their journeys and motivations you can optimise your site to give them the best possible experience. And, as we eventually move back to more business as usual, it also allows you to sell more effectively as we all work together to get the economy back on track.

A key to this is segmenting your audience by behaviour using machine learning. But how do you go about this, particularly when there’s so much else on your agenda?

The good news is that it is possible using Google Analytics.

Using Google Analytics to segment

Google Analytics and audience segmentation haven’t always gone hand-in-hand.

Google Analytics is mainly used to report on aggregate user data and not for looking at individual users – for example, where an individual landed within your website and how they behaved.

Unfortunately, this is the kind of data that you want if you want to segment your audience by behaviour using machine learning.

All you need to do is extract data from Google Analytics using a dimension called the ClientID. This is a unique identifier for each user. Though you can’t combine it with all the other dimensions that Google Analytics tracks, you can access the following data for each individual user:

  • How many sessions they have on the website
  • What channel they were referred by
  • How many transactions they made
  • How much revenue was generated by those transactions
  • How many times they used the search functionality
  • What they searched for

Ready to learn

Once you have this data, you can use a simple unsupervised machine learning algorithm to cluster users by how they behaved.

An unsupervised machine learning algorithm basically finds patterns in data – it doesn’t use any labels to figure anything out, but just clusters based on how similar things are.

You can cluster together users that had a high average order value, or transacted really quickly and find out where they came from, what page they landed on, and what they commonly search for in your internal site search.

As you can imagine, this is extremely useful for segmenting your audience by real-world behaviour. With this, you can identify potential pain points for particular types of customers and improve the intelligence behind your A/B testing programme.

 

 

Real benefits

Here’s how it works in the real world. Recently, we uncovered a segment of users for a client that represented the largest proportion of revenue, visited the site most often and transacted more than once in a 10 day period. However, they had the fewest page views per visit (44% fewer than average) and the least amount of time on site (77% less than average). These users obviously knew what they were doing.

As a result, they were interacting with the site search the least, which made sense as they seemed to know what exactly they wanted and where from.  

However, when they did use the site search, 90% of the searches were for one thing – season tickets.

As a result of this, we found that this was a journey that could definitely be added to the test programme for improvement. An easy win for a high value audience.

It’s easy wins like this you could be implementing now to help your users get the experience they need in troubling times.

 

See how machine learning for Google Analytics can quickly help you optimise your platforms.

Get in touch

More like this