Digital Transformation for Banks and Financial Institutions in 12 Steps

12 Steps to Start a Successful Digital Transformation at Banks, Financial Institutions, and Digital Enterprises

Introduction: Why must banks go digital?

Much has been written about the process of digital transformation among banks, financial institutions, and financial services companies. Digitization and digital transformation are among the most hyped concepts of this decade.  Hype aside, those in the banking industry that put off a digital transformation too long, face an existential threat.  82% of consumer banking transactions will likely occur via digital channels within 5 years. With a new version of the Payment Services Directive (PSD2) and other regulations facilitating consumer account portability, it is imperative that banks and FIs begin competing head-to-head with both better-capitalized banks and more agile fintechs on service. The sooner you embark on a digital transformation journey, the sooner you will be capable of competing for consumers on the merits.  We have helped dozens of multi-billion enterprises through their digital transformations and have learned much about what ensures positive outcomes and what completely derails the best-laid plans.  The difference between success and failure is often subtle and almost always avoidable.  We hope you can use this guide to ensure your own digital transformation’s success.

Short Description:

Digital transformation for banks is not a milestone that, after an incredibly long and laborious journey involving hundreds of consultants, many thousands of employee hours, and tens of millions of dollars, one day is suddenly achieved. It is a process and it does not need to be complicated or difficult to begin.  Much less attention has been focused on breaking down this lofty and often intimidating concept into the simple steps that can be used to de-risk the first digital project and make the journey more simple and actionable.

This guide provides a simple set of achievable steps for mid-and large-sized financial and banking institutions around the world to start or accelerate the process of digital transformation. We’ll begin with the most basic concepts, which you can skip if you’re already in the digitization process, although even if you have started, it is always beneficial to go practice the basics to ensure you have the strongest possible foundation for your organization. Your digital transformation initiative, as some of you already know, is a momentary goal that once achieved becomes an opportunity for continuous improvement, no matter how good of a job you’ve done to date. The steps that we’ll go through are as follows:

Step 1: Identify 3-5 Urgent Problems Within Your Organization (that can be solved, wholly or partially, with deeper data analysis)

Step 2: Familiarize Yourself with Cloud Computing

Step 3: Look at Your Data as it is Organized Today (from a readiness and usability perspective)

Step 4: Choose and Define The Biggest Impact Project

Step 5: Choose a Cloud Provider

Step 6: Set up a Cloud Instance

Step 7: Organize, Clean, and Move the Data into the Cloud Account

Step 8: Explore (data, visualizations, out-of-the-box machine learning, etc.)

Step 9: Deploy a Test

Step 10: Deploy in Production

Step 11: Monitor and Confirm Results

Step 12: Document, Iterate, and Expand

No matter what your role, and no matter whether you are personally interested in the matter or have been formally tasked with helping your organization achieve digital transformation success, you can use this guide to make sure that you are starting with the right framework, asking the right questions, and lining up the right internal support.

Digitization vs Digitalization vs Digital Transformation:

A quick definition of terms that are often used interchangeably but which should be clarified and used more strategically is often helpful. The most frequently misused terms are Digitization, Digitalization, and Digital Transformation.

• Digitization, specifically, is the process of taking analog information and making it available in a digital format. For example, if your new application data is currently available only in paper form, you will need to focus on the digitization (scanning and OCRing and databasing of the form data) as the first step in your Digital Transformation strategy.
• Digitalization, the least defined, usually refers to the moving of a process, project, department, or function from a traditional approach to a more modern, perhaps partially automated, approach.
• Digital Transformation, most holistically, refers to the transformation of the entire organization across critical functions to become more competitive, efficient, productive, and profitable due to the successful implementation of their digitalization initiatives

The 12 Steps to Success in Digital Transformation for Banks.

Step 1: Identify 3-5 Urgent Problems Within Your Organization (that can be solved, wholly or partially, with deeper data analysis)

At the core of every single successful digital transformation that we have studied, assisted with, or driven is a big, meaty, definable problem that could be at least partially resolved through a deeper dive — into the data, into the processes, into the surrounding controls, and most importantly, into the problem itself. Choosing the problem is the single most important element in the success or failure of a digital project.

Step 2: Familiarize Yourself with Cloud Computing

Every digital transformation starts with a basic understanding of the pros and cons of cloud computing. Generally, it starts with internal discussions about the pros and cons of the cloud, which hopefully been a focus within your organization for the past few years, but if it hasn’t begun, here are the stakeholders and the reasons why they should care a lot about the digitization of your business:

CEO (Management):

1. Workforce Productivity: back-office efficiencies in which it is common for the same team to handle applications, transactions, and growth-based processes in 2-4x range, without increasing their stress levels and without forcing them to work 7 days a week.
2. Faster Delivery of Value to Customers: Big tech-oriented initiatives will get done faster, more iteratively, and with more certainty.
3. Practicality: Having a small, well-defined test of cloud computing will give the company a first-hand experience from which it can build further as your customers demand more convenience and speed.  If you haven’t begun, you can make an otherwise steep learning curve much easier to climb with one simple project under your belt.
4. Competitive Advantage: If your sector has a lot of slow-movers, being an early-adopter can vault you forward in terms of competitive position.  If there are a lot of competitors that are already in the cloud, then be concerned that you will be displaced or become irrelevant.  For many laggards, an existential event is building.

CFO (Accounting):

In all likelihood, you will not save money moving into the cloud in the first 2 years. During the transitional period, you will want to have one or two relevant systems running in parallel, the first as it is now, and the second system in the cloud. This ensures that if the cloud system is not performing as expected it is intended, then the cloud system can be taken offline to be amended or re-engineered as necessary. In year 2 and beyond, however, the key compelling metrics which should be of primary interest to your financial institution’s CFO are:

1. Technical Cost Reduction in Years 2-3 Ongoing
2. Greater Controls & Efficiencies:  The cloud affords much greater visibility, transparency, and efficiencies for virtually every process and workflow.  By empowering frontline employees to act, and for managers and executives to measure newly available KPIs,  everyone benefits – auditors, executives, and customers alike.
3. Greater Profitability in Years 2-3 Ongoing:  Even now, profitability at companies that have embraced AI (artificial intelligence) and other cloud-based initiatives average twice that of the laggards.  If that isn’t enough to get your CFO’s approval, you should be a little concerned.

CIO & CTO (Technology): 

Odds are these executives will not need convincing, and much has been written about the significant improvements in tech teams’ productivity, but if they do:

1. Technical Agility: Tech initiatives will get done faster, more iteratively, more efficiently, and with more certainty.
2. More Alignment between Customers, Business Owners, and Developers:  The cloud-enabled agile development method embraces an iterative approach of development, discovery, and validation with product owners and customers.
3. Faster Realization of Value Provided by the Tech Team.  Due to the iterative approach, customer value can be delivered much earlier and more frequently than using more traditional methods.

Step 3: Look at your data as it is organized today from a readiness and usability perspective

Look across the functions that fall within your responsibility, and take inventory of:

1. which data sets are available,
2. what form they are the data sets currently in (excel, csv, txt, sql, etc),
3. how comprehensive is each one is (i.e. is it enough data points to help you solve a part of an identified problem),
4. How much work is required to make it ready and usable?
5. Does the data have associated outcomes? Examples include:

• If you are a consumer lender seeking to digitalize your loan application process, make sure your loan repayment data is available (where the defaults are negative outcomes and full repayments are positive outcomes),
• If you are building a digital retail bank looking to digitalize your customer onboarding process,
• If you are in any growth-oriented function, which outcomes are you trying to create more of and which outcomes are you trying to have less of, and are those outcomes are diligently being tracked and accessible for inclusion in your data set?

Often, digital transformation programs get derailed because there is no agreement on what data to use. Choose a set of data that:

1.  is focused on the problem at hand,
2. can be dumped out into a static file,
3. which does not contain sensitive information (social security numbers, dates of birth, etc), as this will make conversations with your security people much easier. If you need to retain names or SSNs, consider hashing them which will retain their uniqueness, but not add any security risk to the project.
4. easily updated so when you see an opportunity to create value, subsequent improvements and iterations can be made with relative ease.

Step 4: Choose and Define The Biggest Impact Project

You have by now had discussions with your colleagues, talked with your bosses and management, gotten a sense of which projects have the best cross-department interest and support, and which problem has enough data, and which data has enough specificity, organization, accessibility to seed an experiment that will have the highest odds of solving a specific problem, revealing something of significant value or identifying a process that can be greatly optimized.

Three examples of projects that generate better-than-average odds of success are:

1. Fraud prevention: obviously, this is going to be our favorite, since, with a modest amount of transactional data and fraud labels, it is easy to test, has an immediate and measurable impact on your organization’s bottom line, and you can create a compelling ROI case for your CEO and CFO to get the project launched.
2. customer segmentation: transitioning from traditional segmentation to a cloud computing-enabled ‘segment of one’ concept, in which your marketing can speak more relevantly to your customers’ specific needs, is a good medium-term project which can get your CMO and marketing team excited, or
3. Customer service chatbots: if you don’t have any access to data, using off-the-shelf software run in the cloud, you can enhance your customers’ journey and experience using chatbots to answer questions and deal with rote and repeatable work.

Step 5: Choose a Cloud Provider

This is an easy one. First, if your organization already has a cloud account, use that cloud provider. If not, first let’s define ‘cloud computing’ and then there are really only a handful of cloud providers that you should consider.

First, you as an enterprise need to know what it means when you’re moving into the ‘cloud’. What you are choosing in a cloud provider is actually an Infrastructure as a Service (IaaS) provider. This service provides your organization with digital technology infrastructure (servers, operating systems, virtual machines, networks, storage, and other infrastructure technologies) on a rental basis. These ‘cloud infrastructure’ providers will give you further access to other platforms and software to enable you to execute your more specific digital transformation processes. More specific terms you will undoubtedly hear are:

Platform as a Service (PaaS) which are the ‘platforms’ used in developing, testing, and maintaining software to support

1. a) specific functions, such as human resources, customer service, fraud prevention or
2. b) cross-functional technologies, such as database management services (DBMS), business intelligence (BI) services, and email, within your organization. At an enterprise level, it is likely that your organization has, or will have, dozens of ‘platforms’ or ‘systems of record’, each of which serves as the authoritative data source for a given function and which is primarily responsible for optimizing a department’s decisions, workflows and actions.

Software as a Service (SaaS) makes the users connect to the applications through the Internet on a subscription basis.

The 3 best general cloud infrastructure providers are:

• Amazon Web Services (AWS): https://aws.amazon.com/
• Microsoft Azure: https://azure.microsoft.com/
• Google Cloud Platform: https://cloud.google.com/

These platforms are easy-to-use for beginners, are inexpensive to start, and ensure that their services can scale with your needs as you become a more sophisticated and demanding cloud user.

On the security front, your company’s Chief Information Security Officer (CISO) will be instrumental in guiding and confirming the security features in your cloud instance. If you don’t have a CISO, start with the cloud providers’ checklists (AWS’s is here: https://d1.awsstatic.com/whitepapers/Security/AWS_Security_Checklist.pdf)

Step 6: Set up Your Cloud Instance (security, access management – management, data science, etc)

If you have access to tech resources, schedule time to discuss the project with your tech lead. Make sure you have prepared, at a minimum, a project summary detailing:

• the problem,
• data needed, its source and its current and desired form,
• team members, roles and responsibilities,
• executive sponsors, and the
• the process that team members will follow to analyze the problem, monitor progress, and
• iterate if and when necessary.

Your technology team is likely stretched beyond their capabilities and 50-80% of the projects that they are involved with will ever be seen through to completion, so make sure you treat their time as productively as humanly possible. Organize your thoughts and strategies as much as you reasonably can so requests of them can be supremely clear and can be completed with no or few change requests once the technology implementation stage has begun.

If you are the tech resource, here is a guide to getting started at AWS:

https://aws.amazon.com/documentation/gettingstarted/

Step 7: Prepare, Organize and Move the Data into the Cloud Instance

It is tempting to just dump out the data and move it into the cloud to begin the more exciting data exploration and value extraction stages but don’t. Spend time making sure the data is comprehensive, consistent, and as complete as possible. Let’s go through each of the three ‘C’s:

• Comprehensive – is the data broad enough to generate a reasonably good answer. If you are in marketing, for instance, and you are building a ‘propensity to buy’ model, you may only need 10-15 attributes (also referred to as variables, columns, or features) to reach an acceptably accurate model. Since people have been purchasing in relatively predictable patterns for many decades, a well-thought-out set of attributes and a relatively small sampling of records (aka rows or observations). For more complicated initiatives, more attributes and more records will be necessary.
• Consistent – if you are pulling data in over long periods of time and from different data sources, it is important to make sure that the values in each of the attributes/columns don’t dramatically change. If you changed the way you record or calculate an attribute, then either a) shorten the time period to only use the most recent method of recording b) normalize the older set of values within that attribute, or c) don’t use the attribute.
• Complete – for each attribute, look to see if most of the values in your data set are populated. If there are a lot of empty or ‘null’ values, see if you can get the data, and if no data exists, consider dropping the attribute. If the values are uniform, especially those having only one value, it will provide little to no help in solving the problem, so either get more granular data or consider dropping it.

Step 8: Explore (data, analytics, and out-of-the-box machine learning)

And now prepare yourself for a long and (ultimately) deeply satisfying process of problem-solving.  Expect a process, if done right, will take you on a roller coaster ride with extreme highs and lows, characterized daily by both excitement and hand-wringing and by exhilaration and exasperation. Never get too confident in your early good findings or too disheartened when you fail to prove your hypothesis or when your findings are disproven. As long as you keep the solving of the problem as your sole focus, keep the means and methods of solving the problem very fluid, and keep the iterations and experiments agile, you are certain to crack the case.

Step 9: Deploy as a Test

So if you have found a solution to your problem, optimization to the process, or a model for your prediction, remember that as few as 10% of solutions (and especially if machine learning is involved) successfully make it out of the lab and into production. As such, incorporate a test phase into your initial project architecture and strategy. This is the phase that you will see your lab results succeed, but more often fail. That’s why it is a crucial part of any successful digital project deployment into production. Once the initiative hits live production, you will have the eyes and attention of many more stakeholders on the results, so if there are problems, you will need to aggressively expose them in this test environment. Work out the kinks, gain as much confidence that your digitalization project will work in production, then make it live, but not before.

If the improvements or processes fail to live up to the lab performance (plus or minus 10%, for instance), then consider kicking it back to the lab to see if there is a problem with:

• your underlying data
• any transformations that were performed on the data,
• any differences in environments that touched or processed the data,
• any differences in the methodologies that might affect the way the data is captured or processed, and/or
• any differences in the ways users interact with the apps, data-capturing processes, and mechanisms.

If you haven’t yet gotten very granular about the end-user experience, now is the time. At earlier points in your project planning, you may have included the concept of ‘first, do no harm’, and if you did, congrats. In the architecture and strategy, do what you can to ensure launching an experiment into production will have the absolute minimum possible impact on current, well-known processes and workflows. And if the end-user is your customer, then that point is amplified by 10 times, so make sure that any of their experiences have been rigorously debugged, have included their inputs and feedback, and that if something breaks, that their default experiences are minimally impacted.

Step 10: Deploy in production

Gameday is upon us and your team believes your improvement, optimization, or model is ready for deployment. Remember that only 10% of lab-approved experiments get deployed successfully. Maybe through your rigorous testing, you have improved your odds of a successful first-time production deployment two to five times (i.e. maybe you’re at 50% odds of success). Make sure the measure(s) of success, which ideally can be boiled down to a single, simple and intuitive key performance indicator (KPI), has been agreed upon earlier in the process by the team, that your reporting of those measurements are ready, then deploy.

Step 11: Monitor and confirm

On the launch of any new digital initiative, no matter how small, look first to do no harm. As we mentioned before, at the earliest points in the project, the architecture should be such that launching an experiment should create the minimum possible impact on current, well-known processes and workflows. That point is amplified by 10 times when your customers’ experiences are potentially impacted. There are myriads of ways to accomplish this.

Step 12: Document, Iterate, and Expand

Finally, you have made it to the documentation and ongoing iteration stage. You have deployed a key digitalization initiative, your results confirmed your hypothesis and now real-world dollars are being saved. You definitely deserve a raise but hold off on that urge. You will be best served by documenting the journey in a step-by-step guide of how you went about developing your hypothesis, defining what a successful completion would look like, getting the resources, organizing and preparing the data, conducting endless experimentation, testing, validating, and deploying your improvement. Allow others to learn from your successes and failures and, in doing so, you will amplify your success and the benefit to the organization a hundred-fold.

Look at your project through the lens of a data journalist (that’s actually a new and exciting job title) and summarize your efforts in story form. Some of the most significant digital transformations were made much more impactful because a summary was written simply, clearly, and visually, and presented in such a way that it was understood by, and inspirational for, everyone in the organization. The immense value that is created through the hard work of completing a digital transformation can be further multiplied by telling the story throughout the organization, and ultimately celebrating it as an important point in the company’s evolutionary history.

Fraud Solutions

The most effective fraud prevention programs use multiple layers of protection. Designing the perfect security measures is impossible, but instituting a multi-faceted strategy will drastically lower the threat. Learn more about how Fraud.net’s global collaborative network, along with our fraud prevention and analytics platform, supports financial institutions.