Data Science Leaders | Episode 30 | 18:03 | December 7, 2021
Get new episodes in your inbox
As compute capability continues to expand, the banking industry is turning more and more to data science to enable better customer experiences.
Use cases have proliferated, from product recommendation engines to predictive customer retention alerts. These innovations can drive real business value, but managing the rollout of process and technology changes always presents interesting challenges.
In this episode, Chun Schiros, SVP, Head of Enterprise Data Science Group at Regions Bank, reveals how her team is leveraging AI solutions to optimize the banking experience. And with insight applicable to data science leaders in any industry, she shares her change management tips for driving adoption of machine learning among data skeptics.
Hello, welcome to another episode of Data Science Leaders Podcast. I am your host David Cole. Today we have Chun Schiros. Chun is the SVP and Head of Enterprise Data Science Group at Regions Bank. How are you doing today, Chun?
Good. Thanks, Dave, for having me.
Today we have a couple of agenda topics. One is on the nose and pretty obvious: banking and data science. You’ll talk a little bit about your experience in the banking world, use cases, and how it might be different.
The second thing we're going to be diving into is change management, more specifically how to win over data science skeptics. That would be talking to those users who may not trust the models and work that a data science leader or the team may be reproducing. You’ll share some tips and tricks there on how to address that.
Why don't we start at the top? First of all, how did you get into banking? When I look at your background, that's certainly not where you started.
That's certainly not what I envisioned, either, when I was growing up. One thing that was consistent as I was growing up was my love for math and science. I actually chose physics as my focus when I was in high school, and an engineering major in college. When I was in college, I was very fascinated by automatic facial recognition. That's what I chose to do my thesis on. Upon graduation, I decided to come to the US to pursue a PhD in the same area: electrical engineering with a focus on digital image processing. In this whole time of my PhD education, it was a combination of engineering and clinical research plus some need for data analysis and analytics.
I took another Master's degree for statistics and probabilities. I can leverage those theories of experimental and hypothesis design in the data that I collected from the engineering model. I can apply that to clinical problems and help the medicine world to answer some of the clinical questions.
I thought about my next step after graduation. I had a roommate at the time that I was serving in BBVA. We were discussing career development and she said, "Chun, what you're doing from a data analytics standpoint is really applicable to a lot of industries." With that, I stepped into the banking world, served at Regions ever since then and have been doing great.
That's great. When I think of the banking world, I think that data science, in some form, and statistics have been around for a pretty long time. You have the whole actuarial industry that has been around for a very long time. When I look at actuaries and data science today, there's a lot of overlap, I would say.
I'm curious. How has data science evolved in banking over the years? What are the types of problems you're solving at Regions?
I would say that the banking industry certainly has been leveraging data and modeling since the get-go. A lot of financial modeling for regulatory purposes and, also, forecasting budgeting purposes, leveraging time-series data and various different global, macro, and economic factors.
Now, as the data world and computing capabilities continue to expand, the leverage of AI and data science is revolutionizing banking services at every step of the customer's journey. This is from client acquisition, to services, to relationship management, to retention of the relationships. AI and machine learning are used in every step from customer analytics, leveraging analytics for channel preference, optimizations, recommendation engines, personalized advice and services, product and services optimizations—you name it. Problem resolutions, omnichannel servicing, risk analytics, fraud detection, and so on. Leveraging data analytics is really everywhere in banking services.
The single view of the customer is a pretty universal theme, right? I think that data science leaders are trying to solve problems throughout various industries.
I know in the banking industry, not from Regions but from my personal experience, it can be frustrating. When you're applying for a home loan—maybe you've had a checking or savings account for many years, maybe you've also been on the investment side—it's not always true that those various groups within large banks are talking to one another. Is that something of a problem in the banking industry as a whole, or do you think it has already been solved by most banks, and I just happen to have an unlucky banking relationship?
I would say the banking industry is absolutely embracing technology and analytics to optimize the customer's experience. Nowadays, customers expect banking and financial product providers to serve them the same way as Amazon, Google, and many of the tech companies, right? They're looking at experience. They're not looking at brands anymore, especially the younger generation.
The banking industry really is staying on top of this in terms of the strategic focus to leverage this, enhancing the customer experience through omnichannel servicing. It’s about “Know Your Customer 360s,” and being able to provide customized, personalized solutions on a real-time basis. It pertains to the context and life stage of the customers, wherever they need it.
In my opinion, banking experience is really getting more intelligent, purposeful, connected, and likely will be invisible in the future, in terms of the brand. AI is definitely enabling it and accelerating the journey.
Banking is traditionally about relationships, right?
If you actually physically walk into a bank, you might have a wealth advisor there. You might know your bank teller, even, but those days are long in the rear view mirror. Where does data science come in when you're trying to forge a tighter relationship?
Like you said, you move a little beyond the brand per se. You're looking more at the overall experience, right? You have a pretty high bar in terms of experience. Like you mentioned, we interact with Amazon every single day. We know what that experience is like. People look at their banking relationship and think it should hit that same bar. How do you quantify great customer experience from a banking perspective?
Absolutely. I'll address it in two separate domains. In the traditional banking experience, especially in a commercial world, it is absolutely a relationship-based experience. We have relationship managers who know our customers in and out. They know our corporate clients and talk to them on a daily basis.
Now in the realm of data and the higher expectations coming on the horizon, there are challenges that our relationship managers face in serving our customers: meeting their needs at the right place and keeping up with all the information about the corporations. As a result, that's where AI solutions are going to be handy and helpful, to really elevate our capability of services to the corporate clients.
I'll give you an example. Regions has identified that need by closely working with our relationship managers and understanding the challenges that they're facing in serving our customers, to deliver that intimate relationship but at the same time, keep up with all the data that it generated. We create an AI-driven platform where there are multiple functionalities embedded into it. This includes recommendation engines for suitable products for our customers, retention alerts to help identify relationships that might suffer attrition risk, as well as potential financial stress in the near term.
The insights and reasons for those recommendations were translated into an intuitive business language. This is such that our relationship managers can take it, plan insightful and very effective calls with our customers, not only to deliver the right insights, but make them valuable business partners. We provide daily insights at our relationship managers’ fingertips, so they can keep up with what’s going on with their clients without having to go through all the sources of data, pulling the data and then planning calls.
So the relationship managers are a great place to start. Historically they have probably not had the results of outputs of models at their fingertips. They've probably relied on the relationship itself, like taking their clients out to lunch and things of that nature. Maybe they haven’t always been terribly data-driven.
I think times have changed and data science can play a pivotal role in augmenting their skills, ability to understand their customers’ needs, and also to position products at the right time in front of them. How do you convince the skeptical relationship manager that data science does have a role?
I think the key is to really be a thought partner with them. Becoming a thought partner means that we need to first understand what exactly is the challenge that they're facing, and how data and machine learning models and AI solutions can solve it.
The first step is to build the right solutions for the right problems. The second one is to drive adoption in the whole process. By high quality and right solutions, I mean to manage through the entire machine learning cycle from the get-go, from understanding the business problems, defining the targets to solve the problems, data preparation, to model valuations, to post-production monitoring.
The quality of the solutions should be gauged and monitored through the entire process. There are many different metrics that we should measure in the whole process. In terms of AI quality, there’s data quality models, conceptual soundness, accuracy of the models, and stability of the models. During the post-production phase, understand how the model continues to perform, including whether there's any data drift on accuracy. Stay informed on how the model is doing to solve the problem that it is designed to solve.
Most of those are things that a data scientist would be able to understand, right, like being able to look at a drift in various accuracy metrics and other model metrics. Those are things that a data scientist can understand, but how do you then convince somebody who is not steeped in data science, that the model is performing well? Any tricks or advice on that?
The first thing is to really create a sense of ownership, starting from the beginning. Take the users with you along the way in terms of design. Plan the roadmap for the products, understanding the business process and how the solution is going to be implemented in the process.
The second one is to start with the MVP, with a limited amount of features. Stand up a user pilot team that can help you validate the product as early as possible during the whole development cycle. Then continue to receive feedback in a continuous fashion. Iteratively update the products by taking feedback before production. Check that not only does it create the right product for the right users and the right problems, but create a sense of ownership for the user as well. They’re invested in the product, just like the rest of the technical teams.
Once the production kicks off, hold regular feedback sessions with focus groups. Identify product champions who leverage the product to the extent of its own potential, and also share stories among all the users, share success stories, share feedback, take the feedback to the technical teams and beyond, into the products for further enhancement.
I would say that it is no longer a process of the data science or technology team to build a product, toss it across the fence for the users, and then they have their own process. It’s a highly embedded interactive process where the users and the developers continuously have conversations to make sure the product is meeting the needs of the users.
That makes sense. It’s great to hear that you have a pilot team that gets involved in the very early stages of the project, all the way to those feedback sessions. I imagine that it’s useful having them involved in that entire journey, even in some of those feedback sessions to help translate what relationship managers might be saying back to the data science team. Hopefully they're in that middle area where they know enough now about data science, where they can act as translators between those who are just getting introduced to product recommendation engines or what have you for the first time
Do you have any other recommendations for other data science leaders out there with regards to change management and working with users in the world of banking?
Always keep your end goals in mind: to leverage AI and machine learning analytics to drive business impact.
As every project kicks off, ask yourself three questions. Why do we need AI to solve this problem, is this the best way to solve the problem? The second question is: how do you leverage AI to build a high quality solution that meets the customers needs? The last one is really, what is the impact of solutions along the way? Continue to demonstrate incremental value.
Perfect. Well it makes a lot of sense. I really appreciate you joining the Data Science Leaders podcast Chun. If people want to reach out to you, can they link up with you on LinkedIn?
Well great! It was a pleasure having you. Have a great rest of your week. Thank you very much.
Thank you so much for having me, Dave.
Data Science Leaders is a podcast for data science teams that are pushing the limits of what machine learning models can do at the world’s most impactful companies.
In each episode, host Dave Cole interviews a leader in data science. We’ll discuss how to build and enable data science teams, create scalable processes, collaborate cross-functionality, communicate with business stakeholders, and more.
Our conversations will be full of real stories, breakthrough strategies, and critical insights—all data points to build your own model for enterprise data science success.