Login

Marketing and Data Science

Interview: Data Analysis Trumps Specialist Advice - How Direct Banks Function

MIR interview with Martin Schmidberger, Fully Authorized Representative at ING-DiBa Germany and responsible for product and target group management

Low interest rates and sluggish economic growth are not exactly ideal conditions for the financial services industry. Almost daily we are confronted with reports of inadequate capital bases, declining earnings, and layoffs at banks. But while many traditional retail banks are struggling with a business downturn, the direct bank market is enjoying steady and respectable growth despite a challenging environment. Dr. Schmidberger, Fully Authorized Representative at ING-DiBa Germany, offers us a glimpse behind the curtains of this direct bank. We will learn how data technology is used so that bank customers are (more) satisfied.

MIR: Today ING-DiBa is the third largest private bank in the highly competitive German retail banking market, despite not having any branches. How is this possible?

Martin Schmidberger: In a market as competitive as the German retail banking market, approximately 2,000 banks are competing for the customers’ favor. To be successful you have to understand the requirements and desires of your customers perfectly and ideally reflect these in your products. Along with our focus on simple and transparent products, we became involved very early in the analysis of customer data. We recognized that the systematic use of data can lead to a better understanding of customer behavior and customer needs and we regard this knowledge as one of our core assets.

 MIR: How do you get to know your customers at a direct bank? What kind of information helps you understand your customers?

Martin Schmidberger: Many of our analytical questions concern our customers’ experience with us as a bank when they use our products: What channels do they employ, what products do they use? What options are used frequently or less frequently? How do customers get information about other products and services? What procedures do they complain about? Which customers recommend us to others? We can learn a lot from the results of the analyses and optimize products accordingly.

MIR: Also, sales management is surely different at a direct bank. How do you sell additional services to customers who are never seen in person?

Martin Schmidberger: Of course, numerous analyses revolve around the topic of sales. Here, we evaluate customer behavior in order to determine purchasing probabilities using predictive modeling. Based on these probabilities, we manage and optimize our sales, both in the acquisition of new customers and in cross-selling. In the process, we rely increasingly on digital channels. Daily we have approximately 400,000 logins to our internet banking service and almost 200,000 accesses from mobile devices – that contrasts with about 20,000 incoming calls. Well over 90% of our customer contacts already occur digitally today.

About Martin Schmidberger
Dr. Martin Schmidberger is a Fully Authorized Representative at ING-DiBa, where he is responsible for product and target group management. Since 1999 he has built up the Customer Analytics department, which focuses on the areas of data warehousing, campaign management and forecasting models. He is currently working with his team on matters relating to “real-time marketing” and the use of innovative algorithms to forecast customer behavior and “machine learning processes”. Martin Schmidberger earned his doctorate in 1997 and has taught at the Goethe University Frankfurt am Main as an instructor and lecturer since 2012.

About ING-DiBa
ING-DiBa AG is a direct bank based in Frankfurt am Main. It is a subsidiary of ING Groep, which is based in the Netherlands and operates internationally. With more than eight million customers, it is one of the largest retail banks in Germany. Its core business areas in retail banking are savings deposits, securities trading, construction financing, consumer loans and checking accounts. The bank does not operate branches, but instead gains insights into customer behavior and needs primarily through the systematic analysis of customer data. In Germany’s largest bank test conducted by the business magazine “€uro”, ING-DiBa was selected as Germany’s most popular bank in 2016 for the tenth time in a row in a survey of more than 180,000 bank customers. More than 3,700 employees work for ING-DiBa at sites in Frankfurt, Hanover, Nuremberg and Berlin.

 

MIR: How has increasing digitalization changed your marketing?

Martin Schmidberger: The digitalization of customer contacts brings two kinds of challenges. On the one hand, we must achieve the transformation from traditional marketing, like from classical advertising mailings, to digital channels such as online banking. Therefore, we need digital forms of advertising whose response rates, for example, do not fall short of the rates for traditional offline media. On the other hand, the digitalization of customer communication entails a significant acceleration of customer interaction. This means that the coordination of advertising campaigns via multiple channels must be quicker and more automated than before. Online and offline advertising messages must be selected simultaneously and blended into a consistent, customized, multi-channel approach.

MIR: Earlier you addressed the optimization and management of sales. What are you doing specifically?

Martin Schmidberger: Our goal is to determine the most appropriate message for each customer individually and in real-time and to display it on the customer’s digital device, for example for internet banking or on the app. We have developed a modern big data system to adapt our selection procedures to digital channels and to deliver service and sales messages to customers. Along with traditional data from customer relationship management, we also consider current data, such as a customer’s immediate financial situation or specific user entry. This enables messages with very high relevance. For example, we can instantly offer a customer whose checking account has just gone into the red a more favorable global credit facility as an alternative to a relatively expensive overdraft facility.

MIR: So your systems transform data into customized recommendations. What does the technology behind such applications look like?

Martin Schmidberger: High standards for the speed of data processing require powerful, modern big data technologies. Once the customer logs into online banking, we have no more than 500 milliseconds to select the suitable sales or service messages. In this half-second, a series of filters and complex selections are executed in real time to deliver the appropriate sales message. A rule engine manages the algorithms for determining the service and sales messages. Subject area specialists can flexibly optimize and expand them at any time without IT support. This flexibility is important to adapt the system to different applications so that future ideas can also be implemented quickly and efficiently.

MIR: Big data technologies are a natural match with major investments. Can you determine their advantages for sales?

Martin Schmidberger: Yes, we can measure that precisely. Our digital sales have become much more efficient and less costly. By increasing digital sales contacts with our customers, we have been able to reduce the circulation of postal mailings by approximately 80%. Cross-selling rates for existing customers increased in this period by more than 20%. This shows that our digital methods help us selling more successfully than in the earlier “paper-world”.

Our task is no longer to select the best customers for a given product, but to find the most suitable product for each customer that makes contact with us.

MIR: Are investments in real-time technologies also paying off?

Martin Schmidberger: With real-time advertising, we are much more successful in reaching customers at the right moment and with relevant topics. These highly individualized digital forms of advertising result in similar and sometimes even higher response rates compared to personalized mailings. To quote just one example: within a few days, a purely digital campaign to sell our “direct custody account” product had a result comparable to a €120,000 printed mailing. The investment for the digital campaign consisted almost exclusively of creating a corresponding banner and setting up a rule designed for the target group.

MIR: So digitalization has radically transformed advertising and sales. Surely the entire company with all its processes had to change as well?

Martin Schmidberger: Yes, we scrutinized many established processes and replaced them with modern, digital implementations. Most importantly we overcame the pre-existing “sales channel silos” through a consistent system of multi-channel management. Previously the planning and selection of postal and digital campaigns was certainly coordinated, but done in separate departments. With our new focus, we now design campaigns across all available channels in an integrated manner. For example, we only send traditional mailings, if a digital and costless contact had not been successful. Instead of mailing lists, we now create selection criteria and real-time rules for our digital channels. We have also restructured our target group management to streamline it for the new opportunities of increasing digital communication.

MIR: Has digitalization had an impact on the methodology of your data analysis or has everything in this area remained as it was before?

Martin Schmidberger: Analytical concerns in the digital world are entirely different. In the traditional “offline” world, such as mailing, we faced an outbound situation. Using statistical models, we selected those customers who best fit a given product offer. This basic framework changes completely because digital sales entail an “inbound situation”. We as a bank do not initiate contact with the customer. Instead, the customer decides whether and when a contact is made. Therefore, our task is no longer to select the best customers for a given product, but to find the most suitable product for each customer that makes contact with us. In this next-best-offer approach, traditional response models must be adjusted methodically.

MIR: So everything is new, even the methodical approaches to customer analysis?

Martin Schmidberger: Of course, many established analytical and forecasting methods continue to be applicable, but they are becoming more complex. Frequently, the choice of the “best” product offer requires a combination of several regression or response models. In addition, in a digital real-time environment, we can take advantage of additional data, such as contact frequency, time, or placement of advertising space that was not available with traditional offline media. This means that we can now link established analytic methods, such as regression analysis, with real-time decisions about contacting customers online. Typical applications are re-targeting or frequency capping for deciding what message should be used and the maximum number a customer should be contacted. In the future, I expect a further blurring of the boundaries between the proven, traditional techniques of the “offline” world and the situational and real-time possibilities of the “online” world.

Once the customer logs into online banking, we have no more than 500 milliseconds to select the suitable sales or service messages.

MIR: How can you achieve these methodological blends?

Martin Schmidberger: We work on a series of innovative developments that fuse old and new techniques. For instance, as already mentioned, regression analysis still plays an important role. Its primary strength lies in the robustness and good interpretability of its results. Naturally, we also use the new processes of “machine learning”, for example with “random forest” algorithms. In various applications, these processes have shown very promising results and a slightly superior forecasting power to regression models. However, it is currently still difficult to integrate such computationally-intensive algorithms into a real-time context.

MIR: But whoever has the best algorithm wins?

Martin Schmidberger: More important than the choice of algorithm, it seems to me, is our ability to create and use an entirely new category of input variables generated by real-time applications and web data. For example, we see that the response rate depends on the time of day. Accordingly, variables like time of day, the amount of time needed to process applications, the terminal device used or navigation behavior are completely new data that was previously not available. My belief is that the systematic exploitation of such data has a greater impact on response optimization than the selection of the algorithms to be used.

MIR: In conclusion, perhaps a look at the bigger picture. You have provided an impressive explanation of how a direct bank benefits from data analysis. In general, how do you assess the outlook for “analytics”?

Martin Schmidberger: The view that marketing is increasingly data-driven is gaining more and more acceptance in everyday business activities. I believe that in a few years, analytics will be a natural part of sales and will be represented in senior management more prominently than to date.

MIR: Does this mean that in future, career opportunities will depend on big data knowledge and analytic skills?

Martin Schmidberger: Analytics today is still a specialized discipline of a few experts. However, I expect that in many professional fields such as product management or web design, analytic knowledge will become much more relevant. In the midst of the current debate about “big data”, great expectations have been raised for an explosive increase in market efficiency, but these expectations are likely to be only partially met. It would seem that analytics still have some maturing to do, including managing these expectations. Still, I believe more than ever that the analytic expertise of companies will play a decisive role in their competitive success.

MIR: Thank you very much for your assessment and informative insights into the data landscape of Ing-DiBa. We wish you continued success in shaping your digital customer relationships.

Authors

Prof. Bernd Skiera conducted the interview in April 2016.