Leveraging Data-Driven Strategies for Better — and More Personalized — Customer Experiences

Data Basics, Scaling AI Rogayeh Tabrizi, Ph.D.

This is a guest post from our friends at Theory+Practice. Theory+Practice is a data science services company at the cutting edge of AI and machine learning that specializes in guiding organizations to make better data-driven decisions, improve processes, and increase internal capabilities.

Almost all consumer-facing companies — from retail and finance to insurance and tourism — recognize that understanding their customers and competing for their financial attention has never been more essential. Vast amounts of data are being accumulated and stored by companies for both marketing purposes and business operations reasons — and I believe it is precisely the ability of organizations to utilize this information effectively and efficiently that will determine their success.

In a 2018 Forbes article, Aleksandr Galkin wrote that the future of retail was “not to compete for a customer with price wars, but to fight for data to make the buyer's experience as personal and unique as possible.” From my work at Theory+Practice as well as personal observations as a customer, I have noticed a real gap between organizations’ desire to implement data-driven strategies and their capacity to utilize data to drive actionable and effective insights that affect these strategies. It is important to educate both executives and practitioners on the “why” and “how to” of data-driven approaches that consider behavioral factors and drivers of observed outcomes.

In other words, to efficiently address any customer’s needs, a deeper approach — which is based on actually understanding why customers make their decisions — is needed. This makes it easier to see if these needs can be best satisfied through hyper-personalization or by reducing information asymmetries (more on this later in the article) and creating avenues to increase trust and reduce physical and emotional barriers to entry. Often, it is not merely an individual’s preferences that inform his or her behavior and subsequent outcomes, but also who they trust and the tastes and preferences of those people. 

customer checking out in a store

Preferences Are Usually Formed Within Groups

We believe we are unique, but frequently our peers impact our behavior and preferences more than we think. For example, the shoes and clothes we buy, the schools we attend, the careers we pursue — all these choices are impacted by the decisions of those around us. One reason for this is that in environments with imperfect information — such as situations with a misalignment between the perceived value and the price of goods/services — individuals tend to seek advice from trusted sources to minimize loss; or they follow the “wisdom of the crowd,” which leads to the law of averages and the formation of social norms.

Enter hyper-personalization, which for example could be personalized product recommendations using unique customer data such as psychographics or real-time engagement with your brand. This segment-of-one approach allows you to optimize whom you target with key messages and offers through the most relevant points along the customer journey. The success of hyper-personalization strategies depends on the desired outcomes as well as the clarity of the assumptions that inform the “how-to” and the methodology behind the modeling and execution.

It is often assumed that hyper-personalization leads to behaviors that come with improved outcomes, such as increased sales and revenues. If that is the case, it is as important to investigate and understand why hyper-personalization results in such changes. Does hyper-personalization increase a sense of loyalty and belonging to a brand or a community because of having received more personalized and appropriate services? Or does it reduce the emotional and cognitive costs of decision-making (especially in environments with asymmetric information and risks) by providing the necessary and sufficient information through efficient and effective use of data?

Identifying the different reasons — the “why” — can help organizations to tailor specific experiences and solutions for users. Additionally, depending on the extent of each of these reasons, hyper-personalization may not be the optimal solution. For example, if information asymmetry and loss aversion is guiding users’ decisions, then providing information that helps close the gap between perceived value and price can produce a similar result or a mild personalization might get the job done. 

Information asymmetry might show itself as having different item classification in a physical store which is different from what is displayed on the retailers website, leading to customer confusion. Loss aversion is the behavioral response of a consumer that is more considered about not getting their perceived value from the purchase which may actually be different from the actual value.  In other words, why personalization results in better performance could be because it indirectly points to a different problem that could be addressed in both an easier or cheaper manner.

Furthermore, what if hyper-personalization is performed badly or improperly due to data limitations or imperfect models used in analysis? This may pigeonhole individuals into products that are suboptimal for them and impact their experience and future participation and engagement with the product or services. For example, a friend searches for something on your phone and, for the next few months, you receive ads that have nothing to do with your needs!

At Theory+Practice, we have found that efficient use of data — and understanding the “why” behind observed behaviors and outcomes to detect net new signals — are key in designing what we call Intelligent Intervention models. The results from those models  are both actionable and able to provide — and point to — business levers that generate lasting outcomes.

Understanding Intentions Is the Key to Hyper-Personalization

Users reveal their preferences in non-trivial ways, and using data to identify and quantify these individuals’ intentions goes beyond writing predictive models. Understanding, choosing and defining features that are fed into AI/ML and deep learning models are as important as the models themselves, if not more so. At the end of the day, how can we create a more personalized experience or product if we don't know what users want and why?

In short, in this trend towards a deeper and more personalized approach to marketing and client-facing engagement, it is essential that companies understand how to identify and extract practically useful and insightful signals from their data. Data leaders need to know the purpose for this new knowledge, and how to continuously develop and improve methods — which can include hyper-personalization — for generating maximum value for the company and their customers.

You May Also Like

AI-Ready Architecture in the Cloud

Read More

Is All AutoML Created Equal?

Read More

Moving Toward a Citizen Data Science Model

Read More

How to Perform Basic ML Serving With Python, Docker, and Kubernetes

Read More