Case study: Deepening customer relationships and growing customer lifetime value with AI-driven analytics

Discover how Five Star Bank, a premier provider of consumer and commercial deposits, loans, treasury services and BaaS, uses TCS Customer Intelligence & Insights™ to deliver AI/ML powered real-time contextual insights to avoid attrition blind spots and drive omnichannel customer personalization.

Learning objectives:
• What prompted FSB to look at a modern analytics solution?
• How the bank selected the right solution?
• What challenges did the bank face and how did they address them?
• What was the business impact?

Transcript:

Raghunand Curpad (00:43):

Banking called Customer Intelligence and Insights. Today I'm here with Abe. He's from Five Star Bank, Abe Heads the Digital Banking Data and BAS for the bank. And today we want to talk about a fantastic story of collaboration, partnership that Five Star Bank and TCS had and the journey that we undertook together. I think over the last couple of days in the various sessions we've talked about data, we've talked about analytics, we've talked about personalization, we've talked about some of the even more advanced use cases around generative AI. So we want to bring some context to a lot of that in terms of what all of that means in terms of an actual live case study and how to actually go about implementing it. So we are really excited to talk about this case study. So Abe, just on initial thoughts, what are some of the drivers, what are some of the background that the bank had before you embarked on this journey?

Abraham Rojo (01:52):

So some of you know that there are two Five Star Banks. There is one in California, one in Western New York. I'm from the Western New York Five Star, which is a legacy over 200 years of history. And this was five different small community bank including a trust formed a 6 billion bank now, right? It's pretty much kind of typical community bank, STAD serving the community. We operate take a lot of pride in what we do in our physical footprint. I joined the bank two years back board has decided start looking at how we can sustain, how they can sustain as well as grow the bank. And throughout the year of 200 plus years, mostly last 20 years or so, a lot of organic growth and we are one of the bank of choice for businesses. So when I looked at this digital journey and my charter was not even data, we had a very small, I won't say very small, we had a pretty marketing insight team part of community bank and they were serving the day-to-day need for data and insight pretty well.

(03:09)

But when I took over the digital, one of my interesting ambition was to expand the Digital Bank capital D, capital B, beyond our physical footprint. And that's when I started looking at, okay, what is our current business? How can we understand the customer behavior and put not a digital first or customer experience first strategy, but put a customer best strategy. And that's where interestingly enough, we started realizing that we have roughly around 178,000 households actively banking with us probably around 10,000 or so businesses, which we think out of that six, 7,000 are pretty profitable business relationship, mostly white cloud service we offer through our bangers arms and through large commercial teams. Then we started looking at some of the pattern from various silos of data points. And my charter was not to build a data warehouse. So if you are sitting here and trying to solve data quality problem or data warehouse problem, this is not a session for you. So when we looked at it, we understood there is a segment out there, they're not really a retail customer, they're not a business customer, we don't even know. And that's where the journey started. We thought, okay, you know what? This is not going to work. We need to bring in data somewhere. It doesn't have to be a data warehouse or a data program. So we just need to understand what this is. That's where we started the journey

(04:52)

And I'll go to that one. And then we found out a phenomenal opportunity. So that's a really good segue into a simple analysis turning into a business case itself and a business opportunity for us. This journey was not a technology project, this investment was not a data investment. This was a business need. A simple two weeks analysis tend to a really good business opportunity in mid 2021.

Raghunand Curpad (05:26):

So thanks for that Dave. So let's a little bit talk about some of the customer segments and you talked about small and medium enterprises, you talked about micro businesses specifically, which sort of fall in between let's say your retail customer focus, which perhaps has been at the forefront of a lot of these personalization and individualization related use cases. And of course when as TIAs we had actually done a presentation, a demo at the same event a couple of years back where we talked about how there needs to be a specific focus towards small and medium enterprises. Also, given that there is now at least some studies have shown that there is a likelihood of about 8% to 10% of SMBs churning out from some of the regional banks on an ongoing basis. So that's one of the underlying study that we took as a basis to say, hey, let's focus on some of the opportunities for helping banks address small and medium businesses. So in that context, could you talk a little bit about what's your view, what's your some of the maybe even strategies in terms of retention or growth with respect to this particular segment?

Abraham Rojo (06:54):

Yeah, absolutely. And first of all, we are putting the image, that's one of my favorite image. So if you look at the small business owner, she's a sole proprietor. She started this business right out of her hobby and her need in her hangar days, she didn't have enough money. She would go to thrift stores and she would talk to her friends, garage sales, that's where she started collecting good things. Then she's turned a hobby, she turned that into a side hustle. She still probably have a full-time job, but this is mostly a side hustle and these are the type of the customers we identified. They're not really a retail customer, they're DBA, Sole props micro businesses, mostly they need little bit of liquidity or little bit of credit to if the inventory doesn't move or if there could be a small handyman shop so they may have a need to fix their truck.

(07:53)

So things like that. So what we found out is they're either charging their business credit card and we are a 6.6 billion bank even we don't have a credit card portfolio. We have a relationship with a managed card services. We sell their card through our branches. But think about a business owner putting a 10 grand into a credit card and either defaulting or paying high interest or she may walk into a branch and ask for a $10,000 loan. So in both case, what you are going to find as a community bank, this is not a true story for Bank of America Chase or a bigger regional bank. I was with M&D for several years prior to that with Chase and Bank of America. So for this customer, a $10,000 is important and most likely she's going to pay it off. And then what happens?

(08:48)

So your higher value customers. So if the exposure, in our case, if the owner's exposure is more than $250,000 loan more than 250, they have an arm assigned. So that relationship is managed. So anyone up to that threshold, that relationship is currently under the branch network and they do a good job. So if, don't get me wrong, if she walks into a branch, she gets that loan, but there it's not managed. So she pays it off and she walks away and half of the bank she does nobody even know, okay, how is she doing from business at that point right now for this segment, for most of you bankers, if you go back and talk to your branch people or your customer facing employees, they're going to tell you that, yeah, we do phenomenal. But then ask the question, how much is the book, what is your balance sheet?

(09:49)

How much deposit you're gathering? More importantly, how are you helping them from a DBA prop to a micro business to a small medium business to a small medium enterprise to a commercial relationship. So if you can take that journey with that customer, with the help of data insight, you name it, it doesn't have to be a data program, you are not only supporting that business, you're not only increasing your shareholders' value, but also you're giving a huge, huge payback to the community. Think about she's soon, she's going to realize, yeah, I got my credit problem solved and my cash management problem solved. More importantly, I have a partner to talk. I have an accounting expert, I have a finance expert, I have a person to tell me about how I can grow the business and businesses growing. All of a sudden she's going to employ somebody. So that's the journey we need to take. So that's the opportunity. We found actually we have roughly around 4,000 customers in that segment. We identified immediately after we put this program and we'll talk more about how program helped, right?

Raghunand Curpad (11:04):

That's a great segue to understanding for example, how we can equip the relationship managers to be better sort of engaged with the SMBs. So some of the things that we're doing to equip them is give them timely nudges in terms of let's say potential next best actions, but also some predictive insights around the cashflow patterns, some predictive insights around potential churn score, sentiment scores, some prediction in terms of the loan book, how likely are they going to default, how likely are they going to pay off their loan and early, et cetera. So all of these are what we would say things that you can do using technology and equip the relationship managers to be better informed, better equipped in their conversations with the SMBs. So talking about the program itself, so what are some of the considerations? What are some of the initial, I would say, milestones that you needed to hit to get this program up and running, both internally within the bank and also some of the considerations that you applied to get this started off the ground?

Abraham Rojo (12:22):

Yeah, absolutely. I mean I think the next trim age, so that journey started right there. We started unlocking the bundle. So we started unlocking the data and that was mostly we didn't need any technology, although you probably won't. Like we didn't have to even bring in a partner or selected technology to unlock the data. So data was already there, pretty decent quality data. And then soon we realized we may not be able to scale it, Excel access, you start writing our own Python script. So one thing I started when I joined Five Star, Five Star didn't, have a software development shop, pretty much a good IT shop infrastructure support. I started bringing in some of my close accountants from other place to start a true software development shop and they're all techies. They would love to put their head down, put their cap, wear their hoodie, and just start coding.

(13:19)

So they started doing that. Then we started writing some Python scripts and things like that. That's when we soon realized there is something out there which we don't even know. And then that was kind of very interesting journey for probably over one and a half months tegu when I started talking to you. And then I have background of working with the large data providers. My previous job I had ran a large data program with the help of Teradata, Informatica, IBM and Snowflake was experimented, put a small data lake on a w s and it could have taken that same journey, but when coming from a large regional bank to a small bank, so some of what is our biggest challenge spend, so I didn't have that kind of a luxury to start a big grand program. And that's why my two things was extremely important for me.

(14:18)

One is understanding the opportunity cost, the earliest and then time to market. And that's where we started looking for if somebody has already done this, why would I reinvent the wheel? And we started looking around and asking everyone numerous calls, many friends, many partners here. So you all do good job on building something you offer best framework, data models, consultants, strategists to strategize. But unfortunately I didn't have that time. This was not a program, this was part of something else immediately. So an opportunity. So I have to kind of quantify the opportunity and then start experimenting it so that I can fail fast. That's where we look for something out the box. And unfortunately we couldn't find anything at this type of a solution. And even they were not, I think we are probably the first one when enterprise wide on this solution, right in the us.

(15:22)

Yes. So mostly look for something out of the box which can also not only give you your traditional insight and analytics, but actually give a good foundation on trusting the next big thing, nothing but which is AI. This is where instead of me building my own data science team, so I thought, okay, what somebody has already done it, if they have back tested it, if they have experience with it, why can I use that? And that's how we kind of selected a partner and that's the journey we took and which Raghu can talk more about what is the underlying technology, how they came up this solution and how this actually do those three things. And then either you'll be successful, you hit a home run or you can fail fast and both are actually important. You don't want to have a data program which is 18 months running and then realize all of a sudden, what do I do with it? If you're successful as a transformation agent, you move to another job. If you're a failure as a transformation agent, you're out. Then the new person coming in on a data program or some kind of three-year program, he or she or they may not have any idea what to do with it.

Raghunand Curpad (16:39):

Thanks Abe. So the way we looked at this was as TCS, we work with a majority of the banks in the US and we also have the advantage of working with a lot of banks, financial institutions across the globe. So one of the things that we were able to bring to the table was an understanding of the understanding of the banking domain models, an understanding of some of the behavior of some of these, say even the machine learning models that we built and the richness of the data that we had access to working with banks across the globe was really something that set us apart in terms of even say the model accuracy, the prediction accuracy, and the richness of these models. I mean if you look at just our customer analytics data model, it comes with almost close to about 150, 160 plus attributes which are tailored for banks, which are tailored for SSBM context, et cetera, which if a lot of banks, if you were to start this journey and try to build it, it's probably going to take you some time.

(17:51)

So what we are able to bring to the table is a lot of these prebuilt and contextualized and one of the things that we talked about was being able to just leverage this out of the box but also have the ability to be flexible in terms of tweaking it, making it more nuanced for your specific business team's needs. I think we spent about two to three months just making sure that while we did bring out the outbox models, we spent these two to three months working with the business teams, the marketing, the sales and servicing teams just to ensure for example, the churn score that we're predicting is it in line with what the business is seeing on a day-to-day basis. And we've sort of built a very nice intuitive interface for business users to be able to look at say a particular prediction score and understand what is the contributing factor that is sort of giving or telling that particular prediction. So we have this intuitive dashboards, business users are very easily able to, in a lot of ways do a self-serve. It's about them being able to just leverage and see the use case and then come back to us and say, Hey, we think that perhaps there are these additional three, four attributes that we should consider, not from let's say the CoBank system, but maybe you should plug into our accounts payable or maybe a credit card transactions and able to bring that.

Abraham Rojo (19:27):

Yeah, that's a really good point what he's bringing up. So like I said, this immediately started giving us this opportunity of growth opportunity we talked about, right? Identifying that segment, understanding their spend behavior, understanding their stickiness, their credit needs, how much balance they keep on top of that, this also actually started adding more value on the side by opportunity to rationalize opportunity to eliminate reports, canned reports. So a lot of operational efficiency gain that was organically achieved by the program was literally an icing on the cake. And then the program itself was paying off by that efficiency gain. And in addition to that, we started seeing, okay, like he said, how can I bring in the transactions? So I'll give you an example. We have, so we expanded by the way, we expanded our digital bank into six states and I started with the help of my team as well as the leadership.

(20:28)

We started a banking as a service division. So I have the p and l responsibility for that now, which is actually growing. And we started expanding our digital bank, SMB lending, partnering with another feck where this particular segment plus up to two 50 lawn, they can actually originate from anywhere. They don't have to walk into a branch. And these opportunities we got, because this out of the box models were highly the right term. So it was already back tested with the millions and probably billions of data attributes and years worth of history. So models were already built. All we had to do was feed our data to those models and see that initial period of behavior and then overlay that with additional data. So we started, I think second phase, we started bringing in cycle data from an external provider, overlaid that over these models and started doing dynamic market segmentation.

(21:34)

That's a great example where we were able to understand, okay, what is next? And that is right now that is actually influencing our product innovation team. And that actually led us to come up with the two new lawn products a term and an LOC. And in that LOC and term, what we found out was we could actually very cautious. We are a very discover bank and I fully support that, although I like a little bit more freedom, but under 50,000 we don't even need a lot of financial statements or financial documentation for a customer to get it as long as they fall under our credit appetite. And this type of opportunity we started seeing when we started feeding the system immediately, and you can do this, I mean if you have a large data science team sitting there in two years, you can probably build even a better one, although he may not, he will agree, but I'm pretty sure he can build because I'm a builder, I like to code and I have no doubt we can build.

(22:44)

But our point was kind of okay, it's not taking a lot of pride in building that best system. It was what's my opportunity cost? It's not my sun cost or it's not my immediate cost or NPV calculation or anything. What is the opportunity cost if I don't do this? I am not even thinking that. What do I get if I do this? My question is always, if I don't do this, what is my opportunity cost? How many of these customers are using Squire? How many of customers are using PayPal? So we actually partnered with a company called Autobooks and some of you know that we have a digital marketplace now, so we are a Q2 bank. So I expanded the digital bank, added a marketplace, and through marketplace we immediately onboarded three or three, I think four fintechs right now including order books. And then there is a hurdle home base, I think another one UX. So we started giving them non-banking products. How do we know that? We know that they were already doing this? These customers, this S segment was already doing this by looking at their transaction instead of human eye, you can do this. So we started using this AI and then the models was already telling us, okay, X, Y, Z SMB is already spending X amount of dollars for paying their vendors, paying their employees. So that's the insight we started seeing immediately, right?

Raghunand Curpad (24:13):

Absolutely. I think the underlying concept is not based on just addressing one point use case and as an example, I think one of the things that we looked at was it's not just about the data, it's not just about the models or it's not even about how good your models are. It actually boils down to what is the business use case that you're actually solving with that. And some of the things that we saw was, for example, relationship managers or even some of the marketing folks, they were able to do a lot of personalized next best offer recommendations, et cetera. But the key thing was how to make this purposeful, how to make this Intentful and how to make it very segment specific in the context of what the bank is trying to do. So for example, you might want to have say a specific nurturing strategy for your high value customers who are whom you want to do a white glove sort of treatment.

(25:15)

So the next best actions or even the way the nudges that you're giving to your relationship manager may be much more nuanced for some of these micro segments as opposed to say another segment where you are more focused in terms of just the retention. You want to make sure that there is some sticky factor outside of let's say the current merchant services that the SMB might be consuming. So some of these are also driven by not just say a point in time situation. I think it's also important to be able to mine patterns based on say a trend that's happening over three to six months. For example, you could monitor a credit score dipping over a period of let's say three to six months or a higher churn score being predicted, increasing over a period of time and being able to act on that in the context of that microsegment, right?

(26:12)

So your relationship manager for example, will get a very specific nudge that, hey look, this particular micro segment, this particular customer is exhibiting these signs, like his sentiment score is perhaps increasing. So you probably are in a much better position to have that conversation about say new products with this customer given the place at which he's in. So a lot of this is very, very nuanced and what we are seeing is it'll only get more advanced, we go forward. And I think some of the use cases that we are looking at, say subsequently down the line are really exciting. I mean, I feel really happy about some of the things that we can now potentially look at given that there are these foundational building blocks that you've already laid in place. Yeah,

Abraham Rojo (27:05):

Absolutely. So we are on our phase two of this journey and S one is immediately to expand that to our credit risk analytics and order only to look for the default risk, but also early payoff risk moving from it's both a risk for a CFO and a credit officer, both a risk, so early payoff and then how we can actually restructure, give them more or less and help them in other areas. So some of these are, we are already seeing those and then and advance, I know that the big word right now is actually chat GPT. So all of you probably have in your phone already or in your iPad running. So that's where and when most of us decision makers at the bank, probably if you go talk to your board for, Hey, I want to start an AI program, even if board is supportive, it's not going to happen for a 10 billion bank or a 15 billion bank or even a 20 billion bank.

(28:07)

It's a big journey, especially this year, but this actually helping us, okay, they're not afraid of what is AI. Two years back when I joined the bank, if I go to my board and say that hey, I want to have data science, I want to hire a scientist, and then they may say, boy, what is this guy talking about? I don't need any science. I need somebody who can sell the load and bring the deposit in. Now they are seeing this and we are telling them that I don't have to go tell them. Now my body is asking, Hey, I heard about this, what do you think? Isn't that something we already do here? That's the starting. You don't have to start the conversation. All you have to do is get the conversation going. And that's a huge advantage we have right now. I don't have to go educate my executive leadership team or my CEO or my board about, okay, what is AI?

Raghunand Curpad (28:59):

And just to talk about some very exciting use cases that we are looking at. So one of the things that we are realizing is a lot of self-serve can be actually enabled for let's say the business user. So one way we are doing that is we are, for example, looking at generative AI. We've talked about some of these in the last two days as well. But our approach to generative AI and using large language models is being very responsible about it. We are deploying it in firewall. So what that means is all the data, the data constructs, the data privacy related aspects are all within the bank's network and within your firewall they don't leave your firewall. But in the same context, you're able to mine and leverage the benefit of some of the conversational aspects in terms of using generative AI, right?

Abraham Rojo (29:54):

So I think we are almost at the time. So that's a really good point is bringing up, so I can probably ask Chat GPT more than what you say, Chat GPT is also listening to your questions. And that question could be your magic sauce as a banger or as a business or a product owner. And that's where this type of solutions will help you asking those questions to AI within your firewall. So the questions itself is not shared, not going outside. And for AI, the questions are the most important thing, not the answers, because AI is going to learn from your questions. That's an adaptive learning. So you don't want to give out your questions unless it's simple thing telling me, okay, what is Regi chat, GPT or Regi chat GPT? Because everyone knows that those questions are okay, but other things you probably don't want to give it out.

(30:50)

Now just touching me upon another thing, I think another reason for selecting TCS was I wanted a single throat to chalk to implement. And like I said, I didn't have a lot of money, so I went for a fixed bid and I told one thing to Raghu and my best friend Manoj here, when we started this journey, I said, you know what? This is what I have. I don't really care how much time you take, how many resources you put in. And I want the players, I want the people who develop the product to implement. I had no intention to take another SI partner. In this case I selected the product provider to implement for their success along with my success. So that partnership for that partnership, thank you. And more importantly, our customers. I thank all of our customers, check out Five Star Bank. Our bass journey is pretty exciting. I'm a big fan of Blue Ocean rather than red ocean. So I love my FinTech partners and fintechs are doing phenomenal job out there. Say together we can do a better tomorrow for unbanked and underbanked communities.

Raghunand Curpad (32:00):

Sure. So thanks for sharing those thoughts. I think we are almost out of time, but maybe we'll see if there's any questions out there. Maybe one question we can pick. Okay. So great.

Abraham Rojo (32:17):

Thank you folks. Have a good one.