Data & AI: How financial institutions are powering true hyper-personalization at scale

Data & AI: How financial institutions are powering true hyper-personalization at scale

Presented by Envestnet Data & Analytics

AI and analytics are powering innovative use cases for financial services and technology enterprises with the potential for real ROI — if they’re executed right. In this VB Spotlight event, learn how to identify use cases, navigate complex data requirements and more!

Register to watch free on-demand.

In the financial services world, data-driven strategies like AI, machine learning, analytics and more are driving hyper-personalization, at scale — which is arguably the key to pushing product development and marketing to new heights in engagement and satisfaction.

“What’s truly transformative is how these technologies have unlocked even deeper layers of hyper-personalization,” says Om Deshmukh, head of data science and innovation at Envestnet Data & Analytics. “Financial services companies have always had a bonanza of data at their fingertips, but now it’s possible to process it at scale, turning it into structured, tagged and enriched data that can accelerate innovative product development and enable targeting all the way down to the individual customer.”

Enabling new use cases

The insights that data-driven machine learning can deliver span an array of use cases, from safe-to-spend notifications to analysis of a consumer’s retirement goals and individualized recommendations. It can target a user who is ready for a home loan, or open to a credit card upsell opportunity.

On a broader level, Envestnet Data & Analytics has been working with clients on a variety of use cases. For instance, detecting and analyzing inflation, identifying the users that are likely to be negatively impacted, and determining which set of users should be given a warning, and which should be given a bit more leeway.

“Think of it as a loan repayment vacation for three months where the FI can proactively tell them, ‘you’ve been a loyal customer with us for 10 years, we think you may be going through a rough patch — would you like to stagger your monthly repayment?’” Deshmukh says. “You’re offering a once-in-a-lifetime opportunity to show your customer how much you appreciate their loyalty.”

Another client wanted to target those customers requiring assistance in finding their footing after the pandemic to determine what sort of promotional offers could be provided. They were able to identify a variety of potential targets, such as users more likely to take a long-delayed vacation, and would find an offer for a credit card with airline points attractive, or users switching to going out to eat rather than ordering in.

Data-driven insights can also let a financial institution compare itself with its own peer groups, or the local or national macroeconomic situations, and set benchmarks. This is crucial for situations like the recent bank fallout, where an FI needs to determine its status quickly and dynamically in real time, and how to course correct if necessary.

Unlocking data-driven innovation

The turning point for this new level of innovation came thanks to the convergence of three factors. First, according to some estimates, approximately 328.77 million terabytes of data are created each day, which means around 120 zettabytes of data will be generated this year. And financial institutions have benefited from the bounty.

Secondly, machine learning models, especially large language models, are no longer just the purview of organizations with the time, money and expertise to invest in developing them. Today, access to pre-trained models has been democratized, making them accessible to any organization.

And the third factor is simply access to the compute power necessary to run these models. Cloud has made computing far more affordable and attainable, so that companies can run models without the worry of computational and cost barriers.

This is where experienced data and AI partners become invaluable, helping leaders decipher their use case objectives while helping to navigate the complex world of data types and availability. This is done by offering mature end-to-end ML systems, a sophisticated engineering setup, access to the diversity and volume of data that’s required to generate personalized insights, and strongly enforced privacy and security measures that ensure customers are comfortable acting on that guidance.

Data diversity and bias

“We’ve been leveraging machine learning to make data-driven products that touch millions of end users’ lives,” Deshmukh says. “We’ve also been building on the diversity of the data that we have, with very systematic, automated checks and balances to uncover bias or irrelevance, or detect anomalies.”

Data diversity also means that machine learning algorithms should have access to a broad array of data sources. Using a limited number of sources that aren’t representative of your user base or use cases can result in substantial bias in the system.

The inferences you draw from the data should be generalizable and acted on with great confidence. One approach is stratified sampling, where data is sampled across many different dimensions — and those dimensions will be specific to each FI. With this strategy, the models that are trained and inferences drawn pull from as diverse a set of data as possible, enhancing generalization capabilities.

Data enrichment is both eliminating the garbage-in, garbage-out problem, since insights are only as valuable or as accurate as the data you’re working with, as well as adding crucial customer context to every transaction. Every step in a consumer’s daily financial journey adds another piece of important information.

For example, they use their debit card for Starbucks, a credit card at the gas station, and their debit card at the ATM to withdraw cash, and each of these transactions can be pulled together to create a detailed customer portrait — and from there, FIs can realize new personalization opportunities, develop truly precise targeting, leverage data-driven lending strategies and more.

To learn more about valuable use cases, navigating the complex world of data types and availability and more, don’t miss this VB Spotlight.

Register now to watch for free!


  • How does your use case inform the data required for your AI training model?
  • How does data diversity and maturity affect your AI initiatives?
  • What kind of data enrichment is needed to ‘feed’ your AI applications?
  • How might large de-identified datasets help increase your AI solution’s predictive power?


  • Joe DeCosmo, CTO & CAO, Enova
  • Nicole Harper, Director of Corporate Strategy, Jack Henry & Associates
  • Om Deshmukh, Head of Data Science and Innovation, Envestnet Data & Analytics
  • Michael Nuñez, Editorial Director, VentureBeat

Source link

Previous post With Spotify’s ‘Jam’ your whole squad becomes the DJ
OpenAI brings fine-tuning to GPT-3.5 Turbo Next post OpenAI’s GPT-4 with vision still has flaws, paper reveals