Customers should know what they want. But we talk them out of some of their requirements in the interest of their business needs, says Adastra's architect and delivery manager Jan Karban
04. 05. 2022
Reading time: 10 minutes
The digitization of banking does not mean that banks will spam their customers with product offers from morning to night. "Advisers need data to engage with clients in real-time, exactly what they need and in a way that suits them. Among other things, they need to know whether to send an email, call the client, or whether they prefer personal contact," Vanda Gomolčáková, product owner and data scientist at Komerční Banka, describes her experience with digitalization in the ADASTRA podcast.
Ivana Karhanová: Does the digitalization of banking mean that banks will spam us with offers from morning till night, that they know everything about us? Or is it all different in the end? Today I will talk about this with Vanda Gomolčáková, who is a product owner and data scientist at Komerční Banka. Hello.
Vanda Gomolčáková: Hello. Thank you for having me.
Ivana Karhanová: Previously, the banker knew the client personally. Now, slowly but surely, data is replacing the work of advisors. Customers no longer have to go to the bank, and communication with them can be done quickly and remotely. What do you currently see as the biggest challenge for the digitalization of banks? Is it processes, legislation, or customers' heads?
Vanda Gomolčáková: I will start with this area because I work with data. In our team, Advanced Data Analytics or Squat Data Analytics, we process client data mainly to provide banking advice. As a result, we give advisors a more robust and deeper understanding of the client that they can use to talk to the client, plan for their financial needs, or figure out what the client can use to improve their financial situation or keep it stable.
Ivana Karhanová: The digitalization of banks is often associated with marketing. The consequence used to be more detailed segmentation. Now we are getting to the point of Segment of One. How exactly does it work?
Vanda Gomolčáková: For example, in the team, we develop various widgets, such as Miclus, which stands for micro-clustering. It contains many different segmentations, from the simplest one, where the client is categorized, for example, by age or education and by much information they provide us, to more complex machine-learning algorithms, where we can segment clients based on their behavioral characteristics. Each client belongs to each segmentation but only to one segment. This creates client segments that are very similar to each other, allowing us to target them much more precisely. For example, there may be four clients in each segment, but sixteen or even one hundred very similar clients may also be sixteen. This allows us to precisely target and personalize our offer to these clients.
Ivana Karhanová: The personalization is then done through what channels? Are we online or, for example, a personal banker communicating with the customer physically?
Vanda Gomolčáková: It can be different channels. So, as a bank, we try to operate omnichannel. This means that we attempt to communicate through all the media that the client allows us to, by providing us with their phone number, which we can use for marketing communication, email address, or arranging a meeting with a banking advisor. However, one thing is the ability to communicate, and the other is the client's preferences, which we also try to find out using the available data. And for example, if we look historically at what channel the client has communicated through and how they responded if they picked up the phone at all, and how they responded or then bought something, we also try to use machine learning algorithms to see if the client wants that channel if it's friendly to them and if they'll respond.
Ivana Karhanová: That means that on the one hand, we have more data about clients, and on the other hand, we have to weigh how we use that data carefully.
Vanda Gomolčáková: Yes, you can say that because there is a lot of data about the client in the bank, but you have to make some sense out of it. If we just run some big machine on all the data, something meaningful may not come out of it, but at the beginning, there should be some idea of what we want to get with the data and how to use the client's knowledge. And then only then look at what data we have available, possibly some new data to get.
Ivana Karhanová: So you are enriching transactions with additional data at Komerční banka. You have a unique tool for that. What kind of data do you work with there? What is the purpose of this tool anyway?
Vanda Gomolčáková: It's one of the other applications on the Big Data platform that I've already talked about. We call this application Trench, an English abbreviation for Transaction Enrichment. Its goal is to enrich transactions. More or less, we use information about those transactions to do that. We know where the payment went from, whom it went to, and what the amount is for each transaction. Still, we also see the date and time, or what the merchant provides on the post-terminal, for example, some text, like "Thank you for the purchase, Albert Praha 4". We can pull data for the client and the bank advisor from all this information we can put together. If we find a label that says "Thank you for shopping at Albert," we know that the customer shopped at the grocery store, and we can tag that transaction as Grocery. This gives us an aggregated view of the customer's financial and transactional behavior. For the bank advisor, this can, in turn, serve as input to analyze the client's behavior and help them prepare some offers for the client.
Ivana Karhanová: I assume that the advisor then sees the client information in some aggregated form, that he doesn't see that the client shopped in Prague 4 in Alberto.
Vanda Gomolčáková: It's as you say. It can be some view of a month that the client spends a portion of their paycheck or some 30 percent on food, 20 percent on housing, 20 percent they put aside for savings, maybe 30 percent more for the kids, etc. And from this information, bank advisors who are trained to give financial advice can conclude that the client should build up more reserves, spend less at other times, take out a more favorable mortgage, and so on.
Ivana Karhanová: How did advisers use to get information about clients? Exclusively from a personal conversation, or did they have other options?
Vanda Gomolčáková: Also from CRM tools, gradually enriched over the years. A long time ago, in some consultant applications, you could only see basic information about the client. Then, of course, there was some identifying information, he could look at what products he had and open his account statement for that month, but he was reading item by item. These CRM applications are now enriched with just those data inputs. Still, we're trying to aggregate them to a level where a bank advisor can look at the screen and immediately evaluate that client in seconds. There are also different segmentations of these clients, possibly information about having or not having children or what area they work in. The next step in approaching this is to put a tool in the bank advisor's hand that will allow them to get to know the client and have a more substantive conversation. But beyond that, the bank advisor learns the essential things from talking to the client.
Ivana Karhanová: In an institution the size of Komerční Banka, how long does it take to implement such a project?
Vanda Gomolčáková: We are creating the Trench application in our team, and we are already working with customers on deliveries. And when it comes to the development itself, it's about the size of the group. If there were more of us, it would be faster. But right now, we've been working on it for about a year to a year and a half, from the beginning of the development of the whole app to the actual content, and we're probably in the second, third, and third quarter. In this app, the English language workflows, the data processing that we had to program, develop the machiner models, the tagging, and other things.
Ivana Karhanová: Now that you've got it in a production environment, have the consultants tested it in reality?
Vanda Gomolčáková: Yes, thank you for that question. I'm glad you asked that. In the autumn, we are starting to pilot the advisory service in the branches that are our main customers. We call it financial coaching internally, where the bank advisor in the app gets some excellent insights about the customer.
Ivana Karhanová: Who is involved in developing the app on the bank's side? Logically, as data scientists, those who have access to the data, you probably see what you could learn from it. Is it also the consultants saying to you: this is what we are missing, and this is what we need?
Vanda Gomolčáková: So the development is mainly in our country, with data scientists and data engineers involved. They regularly communicate with the business part of the project, with colleagues from financial coaching, who, of course, periodically test the app with bank advisors and also with clients because the app will take one day - I assume next year - to be available for clients themselves so that they can read the data and orient themselves in their financial situation.
Ivana Karhanová: That means that you're putting a tool in clients' hands that allows them to have a better, more aggregated overview of their finances without having to talk to a banker about it.
Vanda Gomolčáková: Yes. And they will be able to plan for themselves how to better optimize their financial situation, for example, where to save more or where to invest more. Or maybe adjust their mortgages to make them more profitable and various other things they can deal with at the bank.
Ivana Karhanová: Have you addressed with clients, for example, whether they are comfortable with the bank handling their financial behavior information for them in this way? I know you usually have that data, but I'm not sure how aware clients are that you have it and can be used in this way.
Vanda Gomolcaková: I think about it quite often. You can't avoid it. For some people, it's very personal stuff. The truth is that we have neither the capacity nor the intention to look at every single client. We track data at an aggregate level. For example, some clients know that we have information about their finances and expect us to work with them and help them manage their finances. They tell us so themselves in surveys. Then, of course, some don't like it and want us not to process their data. These clients have the option to withdraw their consent from us under GDPR. But, of course, all the apps I'm talking about here are subject to regulation, and we don't have the power to process client data for marketing purposes without their consent.
Ivana Karhanová: In Trench, you're also detecting client income. That means you're trying to make some prediction. What exactly can you do for the client based on the data? What can you tell them about the client that maybe they don't know or don't want to know?
Vanda Gomolčáková: Some clients tell us their income. For example, if they take out a consumer loan, that's one of the items they have to fill out. And because the loan product doesn't have that many clients with us, we have to find out from them in some other way. That's why we have machine learning models. With one of the models that we developed, we can estimate from the incoming transactions which one is income for the recurring client, that is still roughly the same amount, or that comes from the same counter-account. Because of the financial advice, we now use a different model that can estimate how the client will be in the next thirty days and their account balance. It considers the regular mortgage payment, the more or less regular restaurant payment, the fact that the client goes on holiday at a specific time each year, and only then does the client's salary come in.
Ivana Karhanová: You want to tell clients how much they can spend beyond their everyday habits.
Vanda Gomolčáková: We're going to warn him or keep him at such a limit that he's still okay. He has some margin. And possibly alerting him that if he continues to spend two days before the mortgage is due, he'll be at zero, and you won't be able to make the payment.
Ivana Karhanová: What's the feedback from clients on this? So if you've surveyed them?
Vanda Gomolčáková: I don't know if this has been tested yet. I think my teammates have tried it with clients, but I'm not sure.
Ivana Karhanová: This seems like a cool thing to me from the client's perspective, that this is what can add value, that the bank will be telling me how to avoid some default or credit risk.
Vanda Gomolčáková: If a client is allowed to overdraft, they can get into a negative balance, but there is a risk that they won't be able to make the mortgage payment. For example, this can mean a negative record with credit companies.
Ivana Karhanová: To give business people an idea: why do we need machine learning or artificial intelligence in these models? What is their role?
Vanda Gomolčáková: In banking, I differentiate between machine learning and artificial intelligence. Artificial intelligence doesn't need to feed any data anymore, and then they talk about learning without a teacher. But machine learning and classical predictive models need a teacher.
Ivana Karhanová: I can imagine that you are then the one who tells him what is essential in this, whether it is the village where he lives, age, income?
Vanda Gomolčáková: Yes, but I set the so-called target with predictive models. That means I'm looking for a client who has bought a consumer loan - that's my number one. And next to her, I have clients - zeros - who have never purchased a consumer loan. The algorithm then splits them into clients who have one and those who don't. So it finds that the average age of the client ones is 40, while the average age of the client zeros is 20. This has been used in banks for many years. And it's mainly used to find clients who want to take out a consumer loan or a credit card loan and to find out the propensity of a client to leave. With artificial intelligence, there are other applications such as chatbots to communicate with the client. These can be some simple algorithms that understand some keywords and search the product sheet for answers to the client's questions.
Ivana Karhanová: You try to communicate with clients in real-time, basically an omnichannel. What's the biggest challenge for the bank so that the communication can respond event-based, real-time to how the client is behaving and what they're doing?
Vanda Gomolčáková: From my point of view, I see technology more as a compliment. I often see my team, and I struggle because the platforms or applications don't always behave the way we need them. So we need to work on the stability of those technologies.
Ivana Karhanová: That is, to be able to load more data faster, to be more stable, to pull only the relevant data out of that enormous amount of data.
Vanda Gomolčáková: Yes. And to make sure that there are no outages because if we have taught a client, for example, that we will label their transactions in real-time and then fail to do that for a few days, they are unhappy. So we need to make sure that when we teach a client a standard, we don't have to unteach them or apologize to them later. This applies to data and applications. In terms of new channels, Komerční Banka is now working on the New Digital Bank (NDB), where they will renew the entire core system and all the technologies that the bank runs on, which will allow both more excellent stability of our systems and more excellent capabilities in real-time communication.
Ivana Karhanová: You will simply be replacing the entire legacy back office.
Vanda Gomolčáková: Yes.
Ivana Karhanová: That's a large project.
Vanda Gomolčáková: Yes, it's a large project. I feel that within the next two years, the first clients will switch to these new applications, which will be more modern and faster.
Ivana Karhanová: Are these projects coming from you, the data scientist, or are they coming from the business? How does it get into the project, into production?
Vanda Gomolčáková: That's a good question because it's somewhere between. I think it would be ideal for people from the business to come to the data scientist and tell us what they need to get more clients into this channel and that channel. So it's happening, but on the other hand, there is a gap between the data world and the business world. Business people can understand clients better than data people, but they often have no idea where the data is and what can be done.
Ivana Karhanová: When you get an idea for, let's say, another use of data, how long does it then take to get it to clients? Are we talking about months, or are we talking about years?
Vanda Gomolčáková: It depends, but it's roughly longer months. Because when you have an idea, you still have to work it out in business.
Ivana Karhanová: Verify.
Vanda Gomolčáková: Everything is tested with clients. There are many specialists involved in developing applications that create prototypes that are sometimes not functional at the beginning. They then create functional prototypes, which the clients try out themselves and find out what works for them. Based on this, a final prototype is made, then programmed. We tend to do projects in the mobile or internet banking area. Some are simple, and they can be done in a month when it comes to data applications. However, more complex applications require more time, sometimes up to a year.
Ivana Karhanová: You mentioned one thing already. What is holding you back and primarily at the moment, is technology, or I guess the original legacy systems that are just being written into other processes there. Is there anything else that would help speed up the data and better use it on the bank's side? For example, should it be more integration of business and data?
Vanda Gomolčáková: Absolutely, there is always something to improve about the communication between data people and business people on both sides. The former should try to understand clients better and go to them and ask questions and verify things. But, on the other hand, there is still a distrust of data in some business people.
Ivana Karhanová: I'm just heading there if you don't feel there are still barriers between the data and the business.
Vanda Gomolčáková: Well, you're certainly not far from the truth. I have colleagues who don't have much expertise in data, but they are enthusiastic. They are the best to work with because they invent things, they tell us about anything, and we know they will mainly use it. Others are still suspicious of data and say they learn everything best from clients. That's true, but try asking, say 1,400,000 clients - you just can't. And then there's also the business people who have confidence in the data and ask us to develop something for them, but then it doesn't always get used.
Ivana Karhanová: So I guess that's part of the process, right? When I'm discovering something, sometimes I miss the mark.
Vanda Gomolčáková: Sometimes, the priorities change so quickly that by the time the stuff gets to the business person's desk, the priorities are somewhere else. But now, the business people in the bank should be right there at the point of request, and they should know what they are going to use our solution for because it takes time, and time is money.
Ivana Karhanová: At this point, we agree that banks are replacing advisors, or their detailed personal knowledge of the client, with data. Do you see any other trend emerging anywhere on the horizon in banking in terms of communication with clients?
Vanda Gomolčáková: I would say that this direction will be even more pronounced. Nowadays, more and more clients do not need to go to the bank.
Ivana Karhanová: I don't want to go there either.
Vanda Gomolčáková: I think you are far from the only one. The coronavirus crisis has helped, and it has convinced us that many things we couldn't imagine before can be done online. And this trend will continue. But, of course, some clients will probably never want a digital bank and will still need personal contact. It's not just the older ones, but we have surveys going back about two years where even young people under 30 have said that they're happy to set up a bank account with an online bank or some platform. Still, when they're taking out a mortgage, they want to go to the bank and talk about what it means, how they're going to repay the loan, and the terms. It's such a big step for them that they don't want to do it online. There will probably always be avenues open for certain types of clients or events. You'll be able to do some things online, and for others, you'll want to go to a bank advisor.
Ivana Karhanová: Says Vanda Gomolčáková, who is a product owner and data scientist at Komerční banka. Thanks for coming to talk about data in banking. See you sometime.
Vanda Gomolčáková: Thank you again for having me, and see you soon.