Skip to main content
How to Track Customer Changes in the Pandemic with AI

Nine months into the pandemic, a “new normal” remains elusive. Consumer sentiment and credit scores are up. But in Europe new lockdowns are in force, US reopenings fluctuate on a daily basis, and masses of new and existing customers are embracing a digital lifestyle as never before.

When uncertainty abounds, how can banks and merchants use artificial intelligence to successfully reopen, understand customer changes in the pandemic and engage with customers to meet their changing needs? This two-part post takes a look at how to do so.  

Data Is Changing

At least one thing is certain in the COVID era: data is shifting. AI can help us gain a deeper understanding of these shifts.

For example, FICO’s fraud business has long used collaborative profiles and behavior archetypes, an analytic construct in which customers can be grouped into a finite number of customer behavior clusters. Transaction behaviors of a cluster can be analyzed to determine whether a customer in that group is behaving normally or abnormally. This technique is useful in pinpointing fraudulent transactions; it also can be invaluable in understanding how customer spending behaviors may have changed, such as during the pandemic. If a customer previously made all of his electronics purchases at Apple but most recently bought a computer from a discount online brand, is this purchase indicative of fraud, a reflection of pay cut or job loss, or spending more within means?

Collaborative Profiling Tells a Story

FICO’s collaborative profiling technologies translate each transaction into an event or “word,” with all transactions past and current forming a figurative “story.” These stories translate the individual transaction detail into a set of behaviors typical of customers. Different customers have different statistical loadings in behaviors, and as new transactions add new words to the story, the behaviors that describe the customer change in response.

Importantly, the archetypes that are learned shift with each and every transaction, providing a real-time measure of the current state of the customer and their anticipated behaviors. During the pandemic, when understanding the customer journey is paramount, collaborative profiling offers a glimpse into the unprecedented struggles that customers may be experiencing, as revealed in their complicated customer journeys.

AI Brings Drab Data to Life

Change can be observed by building a collaborative filtering model based on archetypes and transaction behaviors pre-COVID vs. during COVID, and then gaining an understanding of how a customer’s loadings of behavior archetypes are mutating. Insights gained from this exercise can help us decide how to engage a customer differently; change detection algorithms provide a trigger as when to reach out, and the archetype loadings allow an understanding of how best to do so. 

Analogously, through collaborative filtering we can intuitively understand which people have experienced minimal financial impact due to COVID, while others’ is severe. The pandemic hasn’t turned humans into aliens — people are still buying groceries, getting takeout food delivery and paying their credit card bills — but COVID is indeed driving substantial individual customer behavioral changes. How can banks and merchant understand these changes to understand and appropriately engage customers? 

Toward this goal, applying AI to examine changes in transaction behavior can be extremely insightful, bringing meaning, color and life to the typically drab, static data that models mull over, such as credit bureau or masterfile summaries of customer accounts.

In Part 2 of this post I’ll explore the data connectivity and privacy concerns that are critical in understanding customer changes in the pandemic. Keep current with my latest thoughts on AI, machine learning, fraud and more on Twitter @ScottZoldi.

related posts