Artificial Intelligence (AI) was once a buzzword. Today, AI has graduated to form an integral part of business mandates shaping companies around the globe. 

At Tealium, we believe that consented, filtered, and enriched customer data (that works in real-time) is the backbone of any successful AI initiative. It’s why we built Tealium for AI!

This trend rings no different for the world of AI experimentation. AI has made A/B testing and personalization faster and more scalable than ever before. But, without collaborative access to reliable customer data, building better customer experiences can quickly tailspin. 

To ensure AI makes testing easier, not more overwhelming, we spoke to Fred De Todaro, Chief Product Officer at Kameleoon. Fred stresses the importance of supporting AI-driven A/B testing with clean data to facilitate collaborative workflows. Additionally, Fred talked about the crucial role customer data plays in AI-driven A/B testing.

With a Customer Data Platform (CDP), Fred says AI can become the partner you need to scale experimentation across both product and marketing teams towards common goals – known as ‘All-Team Experimentation.’

Past Customer Data Guides Current AI Decisions


Fred emphasizes the importance of historical data in helping AI make sense of the present. This contextual layer, according to Fred, enables AI to inform and refine current experiment strategies. 

“With access to past customer experiments, AI can draw insights from what has been tried before,” says Fred. “This contextual information can be used to spark new ideas from past learnings.”

Customer Context Is Key For AI-Driven Experimentation Workflows


AI with a working knowledge of customer data can help teams identify what is new to test (and what isn’t). 

Fred emphasizes that access to a CDP can be important for teams looking to identify, prioritize, and manage the firehose of experiment ideas AI spits out. While an effective AI integration should increase output, it also needs to help manage scaled volumes and enable cross-team collaboration. 

“If each person has access to what other teams have been doing, it’s a great knowledge base for insights,” says Fred. “Let’s say I want to run an experiment; I can ask an AI, ‘Does that experiment make sense?,’ or ‘Has this experiment been launched by someone else in the last three years?’ With full access to historical data, you gain instant insight into what has already been done in the past. That’s vital context for AI to function properly.”

AI also benefits from historical context when aiding with data analysis. To put it in simple terms, the more data points that are available, the more context is made available for AI queries. And, with more customer context, AI can provide more accurate recommendations for engaging specific audience segments. However, Fred cautions against over-reliance on historical data for AI-driven experimentation, as it can drive potentially erroneous results. 

Real-Time Data Supports AI Responsiveness

Why Real-Time Data Is The Best Data for AI


If historical data sets the stage for effective AI-driven experimentation, real-time data provides the actors with a well-executed play. Fred stresses that, without access to real-time customer data, it becomes difficult for AI to predict human behavior in the present.

“Real-time data is the best fuel for AI because we get more context about current visitor conversion intent. When this live data is combined with historical data collected by a CDP or Cloud Data Warehouse (CDW), it can make AI’s predictions more accurate,” said Fred. “Scoring or data on past visits alone is too outdated for the present – it’s already too late to take action.”

In this respect, access to current customer data points, through CDP solutions like Tealium, enables single current points of truth for AI to draw from. In experimentation, this not only aids tasks like predictive targeting and opportunity detection but also helps make AI capabilities available to everyone regardless of team or tech-savviness. 

Consider, for example, that you have a feature that is slashing your revenue or you have a sample ratio mismatch throwing off results. Would you want your AI to consider this immediately? Or 24 hours later once it’s been logged? (See our blog, What Happens When You Fuel AI With Bad Data?)

AI Aligns Experiments Towards KPIs


Applying data-backed AI to experimentation is an effective way to grow the collaborative capabilities of your team. By integrating data across various teams, AI can facilitate better sharing of available customer knowledge. 

This crucial customer context enables different teams to be aware of each other’s activities and insights, with the ultimate goal of more informed decision-making and a unified approach to experimentation. 

“With access to the right data, AI can help teams align experiments towards intended KPIs,” said Fred. “When AI is provided with the right context, it should be able to help coordinate teams to prevent experiment overlap or overexposure of visitors to experiments.”

AI Provides Guidance to Beginners and Seasoned Experimenters


On one side of the coin, according to Fred, AI can help guide less tech-savvy teams or new team members to excellence. Drawing from available data, both historical and real-time, AI can help guide them in the right direction. 

“Now, with AI, non-technical or inexperienced teams can simply ask; ‘What’s the best way to measure the impact of this test?’ Or, even better, ‘What are the key takeaways I can get from this test?,’” says Fred. “AI can automatically surface the best methodologies to help these teams get more Return on Investment (ROI) from their experiments.”

Additionally, AI provides advanced experimentation programs with analysis that would have historically eaten up many valuable hours. Common analysis applications of AI include proactively highlighting interaction effects, flagging sample ratio mismatches, uncovering undetected opportunities, and suggesting real-time targeting. 

Comprehensive Datasets Enable Opportunity Detection


Every customer is different. So how do you ensure your team has the tools they need to personalize and coordinate every touchpoint to a specific customer?

When you back an AI with a comprehensive CDP, you ensure customers can benefit from automatic opportunity detection. With a correct implementation, AI-backed opportunity detection can be done even for non-conclusive experiments. 

“Providing AI with access to a large set of data points provided by a CDP helps in increasing the likelihood of finding opportunities on sub-segments for non-conclusive experiments,” says Fred. “When AI has this contextual information, it can identify insights from the past and apply them as opportunities for the present.”

Segmentation, Targeting, and Iteration Should Be Steeped In Data


As a general rule, AI’s ability to analyze and segment customer data can improve targeting in experiments. However, the success of this ability is dependent on the volume of data made available to it. 

When more data is available, AI has more data points in which to discern, identify, and group. This increased granularity has the subsequent potential to increase both the impact of the experiments and prevent overlap with existing team efforts. 

“If each person has access to what other teams have been doing, it provides AI with a great knowledge base for insights,” says Fred. “Let’s say I want to run an experiment to target a specific segment of users. I can ask AI whether this experiment makes sense or has already been launched by someone else three years ago. You should be able to gain instant access to all data concerning what has been done in the past and what is being done in the present. Experimental context helps prevent overexposing your visitors.”

“By using AI to forecast the potential outcome of an experiment, you’re essentially asking it to predict all potential behaviors of visitors,” says Fred. “To achieve more accurate proactive forecasts, you need to trick AI into assuming the experiment is live – a task which would rely on thousands, even millions, of data points. The scale of this task is extremely hard to do without a reliable wealth of data to draw from.”

This article features Fred De Todaro, Chief Product Officer at Kameleoon – a recognized Tealium integration partner. Leverage the Tealium and Kameleoon integration to use your AudienceStream audiences or badges, for more personalized digital experiences. Learn more about Tealium’s integration with Kameleoon. 

Post Author

Natasha Lockwood
Natasha is Senior Integrated Marketing Manager at Tealium.

Sign Up for Our Blog

By submitting this form, you agree to Tealium's Terms of Use and Privacy Policy.
Back to Blog

Want a CDP that works with your tech stack?

Talk to a CDP expert and see if Tealium is the right fit to help drive ROI for your business.

Get a Demo