The data-driven decision-making journey in banking

The term data-driven can seem obscure to laymen but it simply describes a decision-making process which involves collecting data, extracting patterns and facts from that data, and utilizing those facts to make inferences that influence decision-making.

Financial institutions are under enormous economic pressure. McKinsey research found that of the top 500 institutions around the world, 54 percent are priced below book value. Hit by reality, banks have tried all kinds of improvements, especially digitization and cost cutting. Though I believe that most of the banks are full speed in their digitization journey, those moves are limited. Those digitization efforts underlie advanced analytics and scale-up modelling. Sharp algorithms have become table stakes and no longer create competitive edge, as they once did. Today what makes a difference is having the finest data for those analytics and modelling engines to crunch. Only companies with the comfortable data maturity are able to display a sustained competitive edge : industrial-scale solutions to exploit data for authentic business insights and vastly improved decision making. But the canvas is as broad as a bank itself. On the one hand, financial institutions need to achieve the speed, agility, and flexibility innate to a fintech. On the other, they must continue managing the scale, security standards, and regulatory requirements of a traditional financial-services enterprise. 

Banks that know where to steer the wheel can uncover new sources of untapped value

Instead of going with a strategy you assess is best, data-driven decision-making is a strategy that uses data to infer for business decisions. Best-in-class analytics kicks off with high-value questions, not data. The questions must precede the analysis of the data and the most salient ones are generated during the strategy process. To guide the discovery process, it is essential to ask ourselves what problem we want to solve and how much value the solution can produce. For instance, does the company want to differentiate itself from its peers by delivering an outstanding customer experience or by boosting its operational efficiency or by innovating rapidly? Given a clearly articulated vision for the business, what data sets can bring that vision to life and generate genuine efficiency gains or competitive edge? So the trick is to work backward to identify the data needed to achieve the targeted business outcomes. Once identified, it is definitely easier to manage the other parts of the data strategy: defining business outcomes, identifying needed analytics, integrating the data, establishing appropriate infrastructure to retrieve the data (think data lake), and adapting the corporate culture to use the data optimally.

Is there an elephant in the room ?

Advanced analytics and modelling is not about solving our biggest problems; it is about solving dozens of small ones that all add up. It can benefit banks wring small improvements out of almost all their everyday activities, beefing up the traditional P&L levers. Banks have gigantic amounts of data scattered through different departments. Pilots to bring together small samples of information can reveal the potential. The smallest advantage can actually make the biggest difference. Especially in operations, I noticed that these techniques can help to redefine processes and shorten them by several steps. It is the beginning of a "zero-ops" mindset where we try to seek operational efficiency through extreme automation of manual tasks.

Many surveys show that data quality is the biggest barrier (ahead of talent and funding) in their analytics and artificial intelligence banking ambitions. But I notice the first pitfall is to address data quality improperly, which leads to waste critical resources (time and money) dealing with mundane issues. Bad data, in turn, breeds mistrust in the data, further slowing efforts to create advantage. The successful recipe is to address quality at the source when data is created and to focus on data that influences action rather than data that only informs.

Seeing data too narrowly as the territory of Finance, IT or the data science organization, not of the entire business is also a mistake. This leads companies to overlook the transformative potential in data and therefore underinvest in the organizational, process, and strategic changes. Similarly, it becomes easy for staff to impute technology for their quality failures to capitalize on data, when the genuine issue is mediocre management. 

The juice is not always worth the squeeze.  

Decision makers love data. Most banks own a rich set of exclusive information (yet sensitive) on their customers. Real-time data numbers, but also text, voice, and images now exist for literally every action that customers make, every product that banks sell, and every process that banks use to deliver those products. Until recently, organizations have primarily relied on structured data, highly organized data sets that are easy to analyze. Unstructured data, like emails and scanned documents, do not usually adhere to a predefined model or fit into relational databases. The easy path is to ignore it utterly due to the inherent difficulties of analyzing it. Don't you think then that we may miss some of the relevant information in making crucial business decisions ? In addition to the sheer volume of unstructured data, I find it particularly valuable due to its ability to provide rich, detailed and qualitative insight into what is actually happening within the organization. I believe advances in Natural Languages Processing, pattern recognition and cognitive analytics are an absolute game-changer that can empower businesses across industries to collect and make use of unstructured insights.

We should dare also to use alternative data. The data that we have inside our corporation is historical. It is about us. But it is not about the world that we need to understand. Almost everything can be tested, measured and improved and this is truly bringing about a quiet but fundamental cultural transformation in how we make decisions. It would mean that data that exists outside our corporation is far more valuable than the data that exists inside our corporation. Dennis de Reus, Head of Artificial Intelligence at ABN AMRO goes even further in leveraging unstructured and external data : “I expect we will see a lot more applications of AI within the existing processes. Looking further out, I think the role of external unstructured data will grow, for example, by taking into account recent news articles to prevent fraud or other criminal behavior by clients. In the future, AI will support every customer interaction. It will connect you to the best person, it will make personalized suggestions to ABN AMRO employees on the best answers for your question, it will suggest personalized/bespoke solutions, it will automatically summarize your conversation and send you the agreed actions, etc. When you call the bank, you’ll still speak to a human, but they’ll be supercharged through AI.” 

Don't mess with my data !

What many people tend to forget is that for Financial Institutions data is first a license to operate, then it is a basis for business decisions and insights for new business models. We cannot use data for commercial reasons. Not only clients expect trust (in their money as well as their data) from their banks but there are regulations banks have to comply with for data handling (e.g. GDPR, PDPA). So the data that is used is for duty of care (things are done on the benefit of the client). In addition, there is a special attention to fairness and removing unwanted biases. This is something that Malou van Berg as ABN AMRO expert lead in data science integrated very well in detecting financial crime: "When AI techniques are to be used to identify clients suspected of criminal activity, it must first be shown that this AI treats all clients fairly with respect to sensitive characteristics (such as where they were born)”.

Build a centralised data backbone

Deploying scale-up data capabilities across the organisation requires a scalable, resilient, and adaptable set of core-technology components. Simple data management is essential to enable the organisation to leverage data from the bottom up, democratizing data use across teams and roles. The solution is to have a central (yet controlled) place where data can be accessed and used, all in a user-friendly interface that is not just meant for data scientists or geeks. Ideally, data access and wrangling is not just happening in a one-off ETL (extract, transform, load) tool - it must be incorporated into downstreams systems. My experience on a successful journey : Envision a tech-forward strategy, upgrade your core banking systems, rationalize your sources, put in place a modern API architecture, build a mix of data lake / data warehouse and make use of the cloud.

Once the needs and goals have been identified, it is useful to spend some time researching existing software tools. Depending on the situation, it may be more efficient to partner with a vendor than to develop a new solution in-house. Outsourcing often results in better resources, compliance, flexibility and expertise; on the flip side, doing the development in-house usually means better privacy, increased employee goodwill and more control over communication and management.   

Loops beat lines every time. 

There is a need to have a connection between how the data tools are designed and how managers use or apply them. The models and dashboards used complement then existing processes. So it can be easily comprehended by frontline managers. It naturally nurtures a data-driven organization that encourages and rewards proven methodologies and results. Banks need an analytics-ready mind-set, a data governance and a lean methodology for implementation. Does the Front Office really feel the owner of the data ? Using feedback loops enables to be faster to market but also produce better products. Eventually, machines learn just as we do: by trial and error. We want people to use our new tools. Adoption is key. Most of the staff cannot read code or understand the output of a model. To act on these insights, they need smart dashboards with an attractive package that catches the eye. It helps them make decisions and test potential scenarios. Lots of communication and incentives are also necessary to get decision makers to use the new tools. Usually, when they do not, it is because they do not trust what they regard as a black box, fear the impact it could have on their roles, or simply are reluctant to change.

Creating models and analytics is like putting nitroglycerin in your engine. At the end of the day, if the driver does not develop the skills required to race faster, the endeavor is depleted. The skills banks need to make analytics work cannot be contained within a single person or a vendor. Your teams must include true experts on data science, engineering, data architecture, and design. Faking it with people who do a little bit of everything will not work. If you cannot find the right talents, I love the motto from Laurence Liew at AI Singapore : “Grow your own timber” - there are many professionals passionate and keen to re-skill and learn AI, ML, Python, R, Big Data, Cloud, etc even within our own organization.

Scaling up, becoming a product or infrastructure sourcing factory

Many banks and fintechs are engaged in a struggle over the customer-facing front end. But traditional institutions can generate significant value by leveraging back-end assets to create and provide products or services to smaller banks and other businesses. That is because many small and nontraditional institutions are short of core banking products, infrastructure, capital assets, or even banking licenses, and do not have the reach or resources to acquire them. Large financial institutions can address this need by developing a portfolio of white-label products to sell to or through third parties, providing infrastructure as a service. A nice example is BUX, the European broker, which makes use of BaaS - Bank as a Service (individual blockchain bank account, stocks order, clearing and custody services) of ABN AMRO Clearing.

Yet, it would be honest to recognize that the financial impact from even several great analytics efforts is often insignificant for the enterprise P&L. If we cannot scale it up, it means, at best, it will be only a sideline to the traditional businesses of financing, investments, and transactions. In my view, analytics and modelling can involve much more than just a set of discrete projects. In 2019, BCG found that only 27% of companies had reached the advanced stage of data maturity, which they calculate based on seven elements: vision, use cases, analytics, data governance, data infrastructure, data ecosystem, and change management. If banks flex their muscle into analytics and modelling, it can and should become a true business discipline and generate eye-popping returns on investment. It can become a new discipline radically reshaping the old patterns of work and we will see that data monetization correlates with banking industry-leading performance.

References :

1. https://www.mckinsey.com/industries/financial-services/our-insights/analytics-in-banking-time-to-realize-the-value
2. https://www.forbes.com/sites/forbestechcouncil/2019/01/29/the-80-blind-spot-are-you-ignoring-unstructured-organizational-data/?sh=1d6c3219211c
3. https://thenextweb.com/future-of-finance/2020/09/29/5-tech-trends-that-will-redefine-finance-in-the-next-5-years/
4. https://thenextweb.com/future-of-finance/2020/12/07/how-banks-use-ai-to-catch-criminals-and-detect-bias/
5. https://sloanreview.mit.edu/article/getting-serious-about-data-and-data-science/
6. https://www.linkedin.com/pulse/want-ai-team-dont-recruit-grow-your-own-laurence-liew/
7. https://www.bcg.com/en-ch/publications/2020/how-data-can-create-competitive-advantage