Monday, 31 October 2016

How to use Segmentation & A/B Testing in Google Analytics to Generate Better Results



Google Analytics, a brilliant digital marketing tool

Google Analytics is, without any doubt, one of the most used piece of software for companies active on Internet through a website or any web platform. Developed by the successful Californian company, it aims at providing a better understanding of user’s identity and behaviors in order to take actions fostering the digital Holy Grail metric, also known as “conversion rate”.
In our multi-connected era, it is critical for firms to collect and use those data to track their website’s traffic in order to be as effective as possible. Therefore, Google Analytics helps worldwide digital marketing teams to answer both simple and complex questions such as:
  • How many people visit my website?
  • Where do my visitors live?
  • Who are my customers?
  • Which pages of my website are the most popular?
  • What is my conversion rate?

There are hundreds of problems that can be answered thanks to data collected in Google Analytics; but website owners cannot only look at the overall numbers without properly analyzing them.

Source: http://startupstockphotos.com

A/B testing and segmentation, two different concepts linked for digging into data

In this article, Shane Barker – a Digital Marketing Consultant specialized in SEO - explains two different concepts used in digital marketing: “A/B testing” and “Segmentation”.

The first of those two concepts is basically an experiment comparing two (or more) versions of a same web page (version A and version B and so on). Thus, having those different versions will help decision makers finding which one increases the traffic on and gives a higher rate of conversion.

The second concept is about how to divide the website audience. Segmentation is the process of grouping prospective buyers that have common needs or characteristics into segments.
While taking its roots in traditional marketing, this method is still particularly important in the digital world. Cookies, Google accounts, IP addresses for instance are providing Google Analytics with precise and relevant information about a website’s users. Those data can then be used in order to lead precise A/B test operations by focusing on people who are the more relevant for a given conversion and for offering them personalized remarketing campaigns afterwards.

Here are some examples of potentially relevant segment that can be made using Google’s tool:
  • Visitor Type (New visitor vs Returning Visitor)
  • Location
  •  Content Viewed
  • Landing Page Type (combination of on-site and off-site customer behavior)
  • Action taken (which customer have completed conversion goals)
  • Value
  • Demographics
  • Engagement
  • Technology platform (visitor’s devices used)

Comparing those segments’ performance is the key for a strong analysis. Linking those two concepts is a must-do in order to have a clear vision of the reaction of a category of user to the changes made. One shouldn’t take data displayed for granted without digging into details: focusing on a certain population is as important as focusing on certain pages of your website. Creating segments for each test variation may seems time consuming but the relevance of the result will, for sure be higher. While a lot of A/B tests undertaken are leading to no changes in conversion rate, working with qualitative samples is one of the paths toward better understanding of a company’s business objectives.  

Personal analysis

This article wittily highlights the importance of segmentation in order to enhance website (or app) optimization. Nevertheless, the missing part is, based on our knowledge of digital marketing, the statistical relation between A/B testing and segmentation. Qualitative segmentation is important and relevant for solid results. Nevertheless, quantitative relevance shouldn’t be forgotten. The concept linking mathematics and accurate results leading to good operational and strategic decision is called “statistical significance”. Defined by Investopedia as “the likelihood that a relationship between two or more variables is caused by something other than random chance”, statistical significance is, in other words, reducing the risk of inaccurate results by excluding potential false positive data. In order to get enough statistical significance in the test results, the segments made must contain a solid number of users. To do so, marketers can either use statistics calculations* or the tool developed by Evan Miller and available here: http://bit.ly/1rUSvW2

Moreover, this year, Google is launching a powerful suit called Google Analytics 360 (still in beta test version as this article is written).  One of the 6 tools developed by the company will be “Google Optimize 360”, an interesting optimization and A/B test tool that offer direct integration with Google Analytics. With this new update comes two major news for our subject.
The first is a new metric named “Session Quality score”. This indicator, by displaying signs of conversion intentions based on quantitative information such as session duration or number of pages visited, will be a relevant for A/B testing. By using Machine Learning to “predict the likelihood of a visitor making a transaction on your site or app.” (Google), it will enable new opportunities in user testing (and remarketing).
Secondly, segments already used in Google Analytics will now directly be available as targets for tests. This integration in the Google ecosystem will bring a huge time gain and productivity rise.

The last piece of advice


Mixing quantitative and qualitative relevancy while working with data is a key for accurate results. While powerful tools as Google Analytics, Optimizely and Optimize 360 are at the disposal of digital marketing teams; one shouldn’t only rely on pre-made tools and by-default processes. The path leading to more conversion is large and complex but an analysis mind and a great customer understanding will be your best ally in order to catch with your business objectives!



* Statistic calculation is the following: 
More details here



Sijelmassi Mehdi

No comments:

Post a Comment