Banks have significant opportunities to better harness customer information for revenue-enhancing decisions, both in marketing and risk management.
For at least 15 years, ever since the proliferation of data warehouse technology, retail banks have been honing the use of in-depth information on their core customers. The list of benefits is long and includes customer segmentation and targeting; product development and pricing; sales productivity; and underwriting and risk management.
Now, however, further progress is being challenged by accelerated growth in the volume and complexity of customer data that banks must digest. In the new era of “360°data,” electronic channels allow banks to track the full range of customer activities, which offers which offers an exciting new world of possibilities, yet has overwhelmed the current state of the art in managing and applying information.
The reasons are not mysterious. For starters, there is a wider variety of salient activity to track as ever more consumers transition from physical financial services to virtual. People are downplaying branches in favor of multi-channel banking, including the Internet, phone, ATM, and increasingly, mobile technologies. They also are favoring electronic media — debit, credit, online bill payments — over cash and paper checks.
These shifts have produced an ocean of new information that is unique, complex and valuable. Yet while cross-sell-hungry banks are under increasing pressure to analyze and synthesize customer behavior across the full scope of the relationship, today’s most analytically rigorous data applications frequently support only specific products or channels, or integrate only limited aspects of relationships.
Banks also struggle to fully understand the detailed patterns of core customer behavior over extended periods of time. A more comprehensive “longitudinal” view of customers provides a better understanding of both current and emerging needs and opportunities. But such a view introduces new challenges. Data overload — the difficulties of storing, organizing, cleansing and analyzing vast volumes of complex information — has overwhelmed the ability of product development and marketing teams to harness 360°data, slowing progress in managing customer relationships more effectively.
In short, rather than swiftly helping banks to manage customer relationships better, 360°data has invariably translated into Slow Data, deterring revenue and cost initiatives at a time of desperate need for performance improvement. At this critical juncture, banks need to rethink their approach and investments in 360°data to ensure its potential is tangibly realized in higher customer revenue and lower customer costs.
To do this, we believe that banks should initially focus their next round of data enhancements in two areas that present the strongest near-term opportunities for tangible performance improvement: channel-informed target marketing and credit decisioning.
Targeting — In a multi-product and multi-channel environment, banks will need precision guidance to deliver the right offers to the right customer segments over the right channels. A serious revenue and profit stream depends on success in this area.
Credit — In underwriting and risk management, banks can paint a much fuller financial picture of core relationship customers than any monoline provider. Insights into ongoing household financial trends — cash flows into, from and among accounts — offer invaluable advantages in effectively parsing risk; managing credit exposure; timely intervention with troubled credits; and in collections.
To prepare for these opportunities, banks should be considering pilot programs that: 1) refine the understanding of how 360°data can be harnessed for tangible performance improvement in marketing and credit decisioning; and 2) identify the attributes of the business system that will be needed to support rollout and long-term momentum, including the infrastructure plan, the analytical plan and delivery channels.
To be sure, the basics of customer analytics have been in place for quite some time, supported by data warehouses, customer-level profitability scores and predictive models of purchase propensity and risk exposure. Banks have also progressed in the field of time-series analysis, including the tracking of various product vintages and sophisticated fraud detection models that analyze customer behaviors to detect abnormal patterns.
But these capabilities are insufficient to meet today’s challenges. One pressing question is how to sell effectively in a multi-channel, virtual environment where there is far less customer interaction in the branch, but far greater volume and diversity of customer interactions across all channels. Today, on average, retail banking customers conduct only one of seven transactions at a branch, with the majority accomplished through alternative channels (Figure 1).
This migration has introduced some serious skews in customer engagement. Novantas research shows that banks are far less effective at developing relationships and selling products to customers who seldom use the branch. For example, banking customers who predominately use the phone have only half the cross-sell penetration of consumer credit as branch-dominant customers. Yet most banks continue to concentrate on the branch as their primary sales channel. Systematic programs to convert service interactions into sales opportunities, though common in the branch for decades, are still relatively rare in call centers, and far less sophisticated.
Such disconnects clearly illustrate why banks have struggled to take full advantage of the 360°data opportunity. Even though there is more “share of wallet” to be gained with core customers in these alternative channels (and at a much lower cost than is inherent in the physical branch network), doing so will require revising and upgrading the existing customer data paradigm.
At a time when banking strategy increasingly centers on consolidating customer relationships, few banks have successfully drawn a meaningful picture of core customer behaviors across multiple products and channels (Figure 2). There is precious little understanding of how changes with one product impact expected risk and profit on other products, and ultimately the lifetime value across the whole customer relationship. To plug these knowledge gaps, a much richer set of linkages is needed among the various data types, including funds movements among products; shopping patterns; channel usage; deposit, purchase and payment transactions; demographics; and external credit use.
Creating such linkages introduces a new challenge in sequencing a customer’s consolidated activity, essentially translating scattered data vignettes into full movies that portray customer lifestyles, attitudes, needs and preferences. By analyzing time-ordered spending and savings patterns, for example, a bank can uncover a world of difference between households with equally thin checking balances. One household may be loading up on consumer electronics and racking up expensive restaurant tabs; a second may be purchasing household appliances; and a third simply is shoving funds into savings.
The first customer profile, more risky, might value a higher-priced revolving credit card; the second might prefer a short-term installment loan; and the third might be receptive to investment products. For major players, the ability to parse such behavioral differences will become a permanent requirement.
In pursuit of new capabilities, however, 360°data often throws up its own set of barriers. Often the technology teams take the lead in developing databases and interpretive software, but their well-intentioned efforts do not always pan out for the development teams. Defined narrowly, a particular set of technology initiatives may be highly progressive, yet still may prove slow, clumsy and expensive for analyzing time-ordered full customer detail for marketing and credit decisioning. Often the full extent of the problem is not known until well after significant time and resources have been invested.
Amid the intricacies of melding and cleansing data feeds from multiple areas of the bank, honing data analytics and building platforms for various communities of users, the slow pace of infrastructure development and the challenges of data architecture often prevail over customer opportunities. And misalignment among product and channel silos retards progress even further.
To avoid getting swamped in 360°data, executive bank management should start with some fundamental questions: “What are the most pressing areas of performance improvement that can be addressed with our trove of customer relationship information, and what are the specific obstacles to realizing them?” Work on specific applications, such as channel-targeted marketing and refined credit decisioning, should progress in tandem with foundational work on customer data and analytics. In this way, 360°data is placed within the context of banking strategy, and not the other way around.
At a time of accelerating customer migration away from the branch to alternative channels, there is both disruption and opportunity for banks. Marketing and sales practices are in for a major overhaul, with long-standing branch-centric practices giving way to an expanded outreach aimed at effective customer interaction through all touch-points. Yet through that pain a new generation of precision initiatives will be unleashed, given the possibilities to electronically customize offers.
The Case for Time-Ordered Data Analysis
Credit card companies have made a science out of tracking customer transactions over time to manage risk. But skilled banks could do so much more, given their multi-faceted interactions with core customers. The key is organizing data in a way that gives insight into the complexity and fluidity of customers’ financial lives, unlocking new possibilities to serve their needs and preferences.
Making decisions with data requires three key steps: collecting the data, organizing the data so it tells a story, and drawing insight from the story. Banks have made significant progress with the first step by building enterprise-wide data warehouses. But it has proved more challenging to organize the data in a way that tells customer stories so that full insight can be drawn.
A key aspect of telling a customer story is sequencing activities in the order that events happened. It is like a film editor taking individual scenes and frames and putting them together so the story makes sense in a movie. Without proper sequencing, explanations as to how or why things happen in the story — perhaps even the meaning of the story itself — are obscured.
In banking, scenes for a customer story include a wide variety of activities:
- Details on purchase and payment transactions done with accounts and payments media provided by the primary bank (i.e., purchased a latte at Starbucks with a debit card).
- Funds flow among accounts (i.e., transferred money from savings to a demand deposit account).
- Third-party bill payments (i.e., paid a phone bill).
- Direct deposits of paychecks and other funds.
- Channel usage (deposit paychecks at an automated teller machine; checks account balances on a smart phone).
- Third-party credit activity and behavior, as reported by credit bureaus.
- Shopping activity (web views and click-streams).
Together these data tell a story about a customer’s lifestyle, attitudes, needs and preferences. But to tell this story fully, the data need to be considered both simultaneously (which activities are happening at the same time) and longitudinally (which activities follow one another). The fact that a customer has a single large deposit to her DDA account provides some information, but knowing she receives large deposits of various amounts at the end of every quarter suggest a more nuanced story — she is perhaps receiving quarterly commission or dividend payments.
Knowing further that she tends to pay only the interest on credit card balances for the last month in most quarters might suggest she has cash-flow challenges between commission payments. Of course, connecting this pattern to simultaneous activity in savings and retirement accounts could give context to risk: no savings and perhaps she is living a commission check away from a debt problem; regular sweeps to savings and/or a 401(k) might suggest a different attitude and lower risk.
Now consider her specific patterns of transactional activity: debit card charges at a local coffee shop every weekday morning, gas from the same gas station once/week, occasional local restaurant charges on her credit card, few airfare and hotel transactions, and modest purchases at national discount stores. She rarely enters a branch and primarily uses the bank’s online bill pay service the first week of every month. She obtains cash at ATMs almost every Friday. We get a story of someone who stays close to home, lives modestly, has irregular income but consistent spending patterns and is comfortable using mainstream technology for banking.
It is the sequencing of these activities that allows us to understand the connections between her funds flow and expenses. The size and types of purchases she makes with various payment vehicles at various times of the week, month, or year allow us to tell her story with more nuance so we can better understand her interests, lifestyle and likely needs. Telling her story with un-sequenced data would be much more challenging and likely would leave much of the story’s context and insight on the cutting-room floor.
Make no mistake, however, this type of in-depth longitudinal analysis is not an abstract exercise; there many tangible applications that will help banks to make better decisions. Consider, for example, how time-ordered information can identify fundamental differences between seemingly “identical” credit card customers, both core customers who have recently borrowed up to the limit on their accounts.
Looking at monthly transaction patterns, it is clear that one individual carefully confines expenditures to the basics, has not drained savings but simply has incurred contingency expenses for appliances and auto repairs. By contrast, the other individual has higher monthly income, as evidenced by higher direct deposits, but is indebted on multiple consumer credit accounts, draining savings, and has racked up a string of checking overdrafts for seemingly frivolous items. Such knowledge can be pivotal in deciding, for example, whether to raise balance limits on cards or approve installment loans.
The importance of 360°data in this transition seems indisputable. Novantas research shows that among retail banking customers, about a fourth are “virtually domiciled,” meaning they seldom visit a branch after opening their initial accounts. An additional 50% of retail customers actively use non-branch channels.
While it is true that customers use the majority of non-branch transactions for simpler things such as obtaining cash, checking current balances on accounts and making payments and deposits, our research shows that people are getting more comfortable with complex transactions such as opening a new account. There is great potential to win new business in the virtual space, but also to lose it to more adept competitors, even those with relatively little local presence.
The upshot is that instead of a supplement to the branch experience, customer analytics now must play a lead role in the multi-channel outreach. The goal is to analytically determine what product to offer to which customers, through which channel and at what time, to maximize expected value.
Some banks are already putting this concept to work in product promotions, both credit and savings. In addition to refining the mix of product offers made to various customer groups, progressive players are also refining the channel selection, guided by an expanded analysis of purchase propensity, channel preference and expected value. In one instance, a bank obtained a 12% lift in expected value following a channel-informed promotional remix of home equity lines of credit, credit cards and certificates of deposit.
To follow up on their cross-sell successes, banks will need a coordinated program to manage multi-faceted household credit relationships. The goal is to refine and expand the pool of eligible borrowers; lower delinquent recidivism; limit loss exposure; and ultimately reduce operating costs across product lines. This is another major application of 360°data that could be especially valuable with core relationship customers.
At many banks today, risk management remains splintered among product silos, leaving one area of the bank to manage a particular household’s auto loan; another area to handle the home equity line of credit; and still another product group to look after the credit card account. This balkanized approach forfeits underwriting and marketing possibilities to offer credit to more households, increases costs and reduces coherence in dealing with customers.
Coming out of the recession, some banks learned the hard way that splintered risk management practices can have serious consequences. In many cases, multiple collections teams from different product areas were left to individually negotiate with the same household, competing with each other as well as other creditors.
Bankers instinctively know that the sum is greater than the parts. And considering all of the information available from core customer relationships, they should be able to discriminate composite household risk far better than their monoline competitors.
To follow through, banks will need a new generation of risk analytics that look across multiple product areas. This is essential in developing proactive treatments that will help the bank to pursue workout and collections activities in a coordinated way that works best for both lender and borrower.
Underlying customer 360°data analytics is the need to array the data to tell a story. This requires knowing what activities a customer does, and in what order — in short, understanding which behaviors lead to later behaviors (Sidebar, “The Case for Time-Ordered Data Analysis”).
For core customers where multiple data feeds are available, adept banks can also do a better job of detecting emerging household financial stresses and, when necessary, limiting further credit extension. If a multi-product default occurs, a multi-product, time-ordered view that captures detailed collection activity helps to determine and execute the “next best action” in collections activities, reducing expenses and boosting returns.
For many executives, 360°data is an esoteric concept that seems so overwhelming as to be removed from reality. Yet to derive real value from the explosion of customer information, development teams will need to agree on a series of pragmatic steps.
Common developmental techniques, such as best practice surveys or performance benchmarking, will be less useful on this frontier. Even assuming that groups of competitors will share best practices, the field is too new and fast-moving to yield much tangible guidance. Mapping 360°data initiatives to comparable performance results across a competitive field is an equally murky research proposition.
The backward step, in our view, is to accept the default position of incremental progress pursued by separate teams. This often is how technology-led projects get mired in operational details. With 360°data, for example, there is an entire body of thought devoted to issues such as data volume, velocity, variety, variability and complexity. Skilled teams can get lost in this forest and never come back out.
Pilot programs offer a better approach, in that they marry development with practical goals (in this case associated with improved customer outcomes). With channel-targeted marketing, for example, the first step is to assemble a team specifically tasked to build out the business case for performance improvement, incremental revenues versus incremental costs, supported by field-tested propositions. This effort becomes the initial roadmap for priority initiatives and near-term performance improvement. As pilots in the field proliferate, successes can be scaled up to drive revenues and productivity.
As concepts further solidify, the development team needs to flesh out the organizational requirements for long-term success, including the infrastructure plan, the analytical plan, and the delivery channel plan:
Infrastructure planning. Pilot programs provide an initial view of the most critical gaps in the bank’s data infrastructure, exposing areas where it is difficult to access data quickly and efficiently, or where it is in a form that is not immediately useful for business applications. In some cases, the bottleneck is that source information is organized and managed by specific product, functional or channel organizations having a narrow agenda. In other cases, data will be confined in a centralized organization that has grown unresponsive to business needs. In still other cases, pilot programs will expose situations where a sufficient history of relevant data are not preserved or, in the extreme, never collected in the first place.
Importantly, these gaps will be exposed within the context of specific programs to enhance revenues or reduce costs. Over time, findings will become the basis for a data strategy that captures and retains all relevant information and makes it available to all of the business lines, in a form that is useful for their needs.
Analytical plan. The assembly of 360°data for pragmatic applications will create new demands for analysis and insight. To make sense of it all, banks will need to create new frameworks, metrics and models. These analyses will include descriptive profiles and predictive models that address the patterns of customer behavior across multiple products and channels, and the evolution of behavior over time. Ultimately, banks will need to understand how all these factors come together to drive Customer Lifetime Value, a measure of performance which will likely become increasingly central to managing the bank on a relationship basis.
Delivery system. Analytical insight has no value if it cannot be translated into action. For that reason, the most critical organizational capabilities may well lie in the operational delivery of new insights.
At most banks today, for example, the ability to drive effective customer dialogue with customers on the phone is surprisingly primitive. Elsewhere, online interactions have the potential to be highly personalized, and yet the ability to exploit that potential is often non-existent. Banks will need to deliver advanced decision science into these channels, and to manage a coherent and consistent personalization logic across multiple channels (including the branch) in order to compete effectively in a world of 360°data.
Capturing Relationship Value
The explosion of 360°data presents banks with a set of opportunities to improve revenue generation and profitability at a time when they desperately need both. More than ever, banks need to capture the full value of the “primary bank” relationship, and win the full customer wallet.
Virtually all product sales, portfolio maintenance and risk management activities are improved by a time-ordered relationship view of customer behavior that transcends individual product, channel, and functional silos. The urgency of realizing these improvements is increasing rapidly. Profit margins continue to be tight. And with each passing month, fewer customers look to the branch, eroding the bank’s traditional approach to reaching them still further.
Banks that win over the next few years will recognize the need to transform their approach to customer analytics, and recognize the new demands of 360°data on their organizations. Along with driving immediate performance gains, well-managed pilots will shape plans for a practical, systematic program to build the next generation of infrastructure, analytic models and delivery capabilities.
Alan Schiffres and Jim Bramlett are Partners in the New York office of Novantas LLC, a management consultancy.