Skip to main content
Understanding data types in A/B testing and personalization

Understanding data types in A/B testing and personalization

March 12, 2021
Reading time: 
4 minutes
Chris Measures
Chris Measures
Chris is responsible for creating Kameleoon content across the areas of personalization, experimentation and digital marketing. He covers a wide range of topics, with the aim of better informing marketers and brands and helping them increase conversions and revenues.

Brands in financial services and healthcare are currently suffering from data paralysis.

While they have access to a huge (and expanding) volume and range of personal data, many struggle to harness it to make better personalization, digital experiences and products.

Forty six percent of banks surveyed in the BAI Banking Outlook report admitted they could make better use of this data to improve product and service recommendations.

The patient experience, including through digital channels, is 5x more likely to influence loyalty, yet only 51 percent of healthcare companies saw using data to improve the patient experience as a priority.

But it isn’t an industry wide problem. Fintech companies grew by an average of 13% in Q1 2020 - while 36% of traditional US banks were downgraded to a negative rating by S&P Global during the same period. Digitally savvy companies can tell the difference between types of personal data, particularly which are covered by consent and compliance regulations, such as HIPAA, GDPR and CCPA. Once a company understands the types of data it has, it can build a strategy for compliance and A/B testing, a key practice to creating better customer experiences and products.

Essentially, customer data comes in two main flavors - hot and cold, and can be further subdivided by whether it is anonymous or personally identifiable.

This article aims to help all brands by explaining the different types of data, their relationship to compliance and how they can be effectively harnessed.

The fear of doing the wrong thing shouldn’t mean a company fails to do the right thing.

1 Hot data: revealing visitor intent

Hot behavioral data is generated by visitors to your website, covering everything they do when browsing. Normally this is anonymous, unless the visitor is logged in and consequently identifiable.

As James McCormick of Forrester points out in the analyst’s “Adopt AI for Personalization Safely and Smartly to Win European Customers” report, “Anonymized data such as visitor website behavioral data is relatively low risk to use” - yet too many approaches deliver personalization that is irrelevant to the consumer, irreverent (being overfamiliar and disrespectful) and irresponsible in how data is collected and used.

Hot data fits into three categories - here’s how to use each of them effectively:

Visitor behavior on your website:

  • What people have people clicked on
  • The frequency of their clicks
  • Where they have come from
  • How much time they have spent on specific pages
  • How their journey has progressed through your site
  • Their history with your site and the number (and timing) of the visits they have made

Visitor data is essential for experimentation programs to build your testing roadmap - for example you can use it to test changes to the user experience in order to ensure they appeal to your audience. Analyzing behavior can also show when a visitor is about to leave, giving the opportunity to trigger an action designed to keep them on the site and moving down the sales funnel. All without using any identifiable personal data.

Boosting loan applications through personalization

For example, one online credit provider increased applications by 8.3%, simply by displaying a personalized message to visitors about to leave its site.

Outline visitor information:

  • Their location i.e. geolocation
  • The type of device they are using (i.e. mobile or desktop)
  • The browser being used

Geolocation gives a second layer of detail on your visitors that you can use. Many banks and pharmacies have already discovered that just by showing the location of the nearest ATM or store, based on geolocation, they can reduce customer service requests. That’s particularly important when it comes to users on mobile devices who want to access particular functions and information when they are on the move.

Contextual information:

  • The weather where the visitor is
  • The day of the week
  • The exact time

You can easily extrapolate information based on the visitor’s location. This allows you to offer a more personalized and engaging experience - for example a pharmacy could highlight hayfever remedies if the pollen count is high in a particular area, or an insurer might offer ski insurance to a visitor from somewhere blanketed in snow.

2 Cold data: enabling deep experiential change

Generally cold data is information you already have on a customer or visitor, provided from other systems such as your core banking platform, CRM, DMP, analytics or CDP. Normally this is personally identifiable information (PII) but some data from DMPs can be anonymous.

Normally this is personally identifiable information (PII) but some data from DMPs can be anonymous.

Cold data could be:


  • A record of how a customer or patient has interacted with you
  • When they last logged into their account
  • The services they have (such as checking account and credit card in the case of banks)
  • The channels they normally use and when they use them
  • How much they have in their account, and when payments come in and out

Personal data from other systems

  • Gender/age
  • Credit history
  • Medical history
  • Memberships, such as of VIP programs

You need to gain flexible consent management to use this data to personalize or experiment, while ensuring you remain compliant - but it can be achieved, and advanced testing platforms make the process clear and straightforward. It is about putting the customer (and their needs) first and building trust that you are actively helping them achieve their goals. That’s why lots of challenger banks successfully cross-sell more products by better understanding their customers and their needs. For example, one online bank we work with increased loan applications by 20% by displaying a personalized offer to logged-in customers, based on their credit history.

Ensuring compliance doesn’t have to lead to data paralysis. The opportunity is there for banks and healthcare providers - they just need to build a better understanding of their data estate and put in place the processes, technology and programs to turn customer data into competitive advantage through a data-driven marketing strategy on your website.

Want to learn more? This is part of an ongoing series on compliance and experimentation. Read our previous blog on what you need to know about A/B testing and personalization in healthcare and financial services. At Kameleoon we’re expanding our coverage of testing, personalization and compliance to help companies get the most from their customer data. Subscribe to our newsletter as our coverage accelerates this spring.

Topics covered by this article
Chris Measures
Chris Measures
Chris is responsible for creating Kameleoon content across the areas of personalization, experimentation and digital marketing. He covers a wide range of topics, with the aim of better informing marketers and brands and helping them increase conversions and revenues.