Business Intelligence: Up the Octane

red formula 1 race car speeding along a track
Contributing Writer

14 minutes

Bad data can cause credit unions’ operational and marketing engines to sputter and stall.

As credit unions rely increasingly on data to power their operations, running on cleaner “fuel” will improve performance.  

“At a fundamental level, credit unions need quality data to understand the markets they’re in and the members they’re serving and want to attract,” says Shazia Manus, chief strategy and business development officer with CUNA Mutual Group’s AdvantEdge Analytics, a CUESolutions provider based in Madison, Wisconsin.

Data quality is at the core of intelligent decision-making that impacts every member service interaction, operational issue and regulatory requirement, Manus says. By extension, the difference between good data and bad tips the scale toward success or failure in identifying and implementing strategic initiatives and in deploying such advanced technology as artificial intelligence and machine learning. 

Bad Data Gums up the Engine

A useful first step toward improving data quality is to define how information can go bad. “In today’s day and age, our decisions are being driven more and more by data, and the idea of bad data itself is changing,” says Karan Bhalla, CEO of CUES Supplier member CU Rise Analytics, Vienna, Virginia. “Data can be just flat-out wrong. Or it can be stale, which is a problem if there’s more recent information out there.”

Incomplete data is another concern, and one that is especially common and problematic for credit unions, Bhalla suggests. He offers collections as an example where only a partial view of member relationships can be troublesome. Especially if different areas of collections such as credit cards and auto loans are outsourced to separate third parties, the result can be a siloed perception of members’ standing. A vendor collecting past-due credit card payments will apply the same tactics to all members on its list, even those who are current on their auto loan payments and have multiple active deposit accounts. Equipped with a more complete view, a credit union would likely take a different approach to collections for members with deeper relationships.  Given the high potential for inaccurate, outdated or incomplete information to send credit unions down the wrong path, Bhalla adds a word of caution: “Bad data happens more often than we think,” he says. “Everybody is having a lot more information thrown at them every day. Managing all that data requires a big culture shift and new expertise.”

Shazia Manus
Chief Strategy and Business Development Officer
AdvantEdge Analytics
Standardization in defining data is a huge issue, and this can be especially hard to overcome when data ownership and stewardship are not clearly defined.

Warning Lights Across Lending and Compliance

As credit unions’ primary source of revenue, a strong lending program relies heavily on accurate, timely information. Basing decisions on incorrect or partial data could lead to denying loans to creditworthy members or approving loans that are more likely to default. Both of those paths can significantly impact a CU’s financial performance.

Regulatory compliance and reputational risks exacerbated by problematic data can also arise throughout the life of a loan, Bhalla notes. If a credit union raises a member’s credit card interest rate because of late payments, the CU is required to check on a periodic basis whether that member’s payment history is back on track and, if so, to reduce the rate accordingly. Both the initial APR adjustment and subsequent actions must be based on sound and complete information.

Inaccurate or incomplete, siloed data can interfere with the ability to comply with a wide range of regulations, Manus says, from anti-money laundering reporting requirements to restrictions on outbound calls under the Telephone Consumer Protection Act. 

“If information is not properly organized, curated, standardized and accessible to business units, the credit union is not able to act in an intelligent way,” she adds. “That increases hidden costs to capture data and remediate it.”

And the longer credit unions have been in business, the more problematic bad data may be, as inaccurate or dated information aggregates in disconnected systems, Manus cautions. Those issues can be exacerbated by poor data governance and processes, ineffective tactical solutions and tools, and weak rules governing data systems and storage.

Incomplete or inaccessible data could also make it difficult for credit unions to apply the current expected credit loss standard set to take effect in 2023, says Carlos Caldera, SVP/chief data officer for CUES Supplier member Trellance, Tampa, Florida. “CECL will require a lot of granular data details. If that level of information is not available, that’s going to impact the amount of provisions the credit union will need to put aside to cover those contingencies.”

And in the IT department, “the lack of good data discipline shows up in the proliferation of data sources,” he notes. “Without good controls about where data is, a credit union can end up with duplicated data, even duplicated databases, and that obviously is going to increase storage and maintenance costs.”

Clogging the Marketing Fuel Pump

Marketing offers many examples of how bad data can increase expense and derail the potential to build business with existing and new members. Without current email addresses and mobile phone numbers, credit unions can’t communicate via members’ channels of choice. Without psychographic data—information about interests, values and personality traits—credit unions can’t develop strategies in line with members’ evolving preferences related to financial services. 

Without accurate, complete information, “you will see poor uptake, at best, if you’re trying to target and run a campaign,” Manus says. “At worst, if you don’t know who your members are, what their wallet share is and how they’re engaging with different channels, you’re likely to see higher rates of attrition.”

When marketing credit cards, for example, managers should be able to classify members into cardholders who rely on revolving credit versus those who pay off their balances each month and differentiate inactive accounts from occasional and high-volume transactors. 

“If you don’t know who these members are and how they are using credit card or payment channels, you won’t be able to engage them with the right offers for rewards or rates to keep your credit cards top of wallet,” Manus says.

Effective data management and digital maturity allow credit unions to take a wider, more active view of their progress toward growth goals. Leaders of one credit union working with AdvantEdge Analytics were committed to using data to drive business strategy and activation, Manus recalls, but they were confused about why success in signing on new members didn’t have the anticipated impact on financial performance. 

“In working with the data, they discovered that member attrition was actually siphoning off almost as many members as they were gaining each month,” she says. “Having a holistic view of clean and accurate data told the whole story and showed that it was prudent to allocate some funds and focus to member retention strategies alongside attracting new members.”

Time for a Tune-Up

These data management challenges can be especially daunting for credit unions with information stored in diverse systems: core processing, consumer loan origination, mortgage origination and servicing, marketing, and multiple channels for debit and credit card, ACH and direct deposit processing. Committing to the following strategies can help clean up all that information and bring it together to form a clearer view of member relationships.

Set a data management baseline. Improving data management encompasses data quality across such dimensions as accuracy, completeness, timeliness, consistency, integrity and validity; data rationalization, with a focus on controlling duplicates or redundancies; and data architecture, which includes specifications to define data requirements, develop a disciplined data strategy and guide data integration from multiple sources. Caldera recommends conducting a gap analysis to evaluate how far the credit union needs to go to achieve these capabilities and then developing a road map to identify the necessary resources and guide implementation across the organization.  

Whatever organizational structure a credit union adopts for these efforts, it must commit to a culture change in support of data management, he adds. “This is not a one-, two- or three-year initiative, but an ongoing commitment. It is more a change in lifestyle than a project. It is harder going on the front end to implement new practices and systems and train staff to develop the necessary skillsets. But even once you’ve reached maturity, you need to maintain data governance and continue to execute effective practices.”

Make data management everyone’s business. A fundamental message is that data is an enterprise-wide asset and belongs to the entire credit union, not just a single functional area, Manus recommends. 

“That requires a whole different cultural lens to understand the importance of data at the enterprise level and to break down siloes and work together to develop a single view of members and their interactions with the credit union,” she says. “The organization has to have a clean foundation on which data is added every day, with transactions and member interactions through all the different channels. Even something as simple as a change of address must be changed across core, lending and other systems to facilitate marketing, contacts, fraud alerts and other interactions.

“Each of these disparate systems has different data definitions, different standards and different ways information is being ‘data-fied.’ Standardization in defining data is a huge issue, and this can be especially hard to overcome when data ownership and stewardship are not clearly defined,” she adds. 

Introducing new technology alongside legacy business systems can exacerbate these problems. Manus cites the example of a new call center interactive voice response system with the capability to store data from conversations with members. 

“How are these new sources of data being integrated with existing systems and data generation?” she asks. “This brings up the potential for more duplicates and overlap.”

Hone data management expertise and accountability. “IT staff and managers may not have complete training around issues raised by the increasing volume of data coming at them, not just through core and lending systems, but social media and psychometric data that is available,” Bhalla cautions.Larger credit unions are beginning to recognize the need for a chief data officer to lead development and implementation of data management strategy, but at small and mid-sized organizations, those responsibilities may be assigned to the CIO or IT managers, he notes. Their approach to data management may be informed primarily by their experience in programming, tech support or networking, which may limit their ability to make informed decisions about how data from diverse sources is used, managed, distributed and decisioned across an organization. 

Karan Bhalla
CU Rise Analytics
With any kind of a campaign, you should analyze what the level of impact could be—the potential downside of getting it wrong. If that comes out high, that’s an indication that it’s worth the extra time to make sure you’re using good data.

Assess risks and set priorities. An all-too-familiar challenge is the need to make the most of limited resources in an ever-changing environment. In marketing, for example, Bhalla recommends evaluating both the possible return and the relative likelihood that bad data could increase risks in targeting offers.   

“With any kind of a campaign, you should analyze what the level of impact could be—the potential downside of getting it wrong,” he says. “If that comes out high, that’s an indication that it’s worth the extra time to make sure you’re using good data.” 

Consider the difference between two marketing efforts: A credit card upgrade campaign that moves qualifying members from gold to platinum cards has the upside potential to increase transaction volume with limited downside potential for added risk if the offer is extended to a few members who don’t meet the criteria. In comparison, pre-approved offers for home equity lines of credit, with loan amounts running up to $50,000, convey more significant risks. 

“If a credit union offers these loans to just 20 credit-challenged members based on bad data, the potential losses could amount to $1 million,” Bhalla notes.

To improve their capabilities to conduct these types of proactive evaluations, credit unions need to spend more time quantifying the costs of decisions based on inaccurate and/or incomplete information. 

“How often do credit unions perform post-mortems to determine to what extent their charge-offs were a result of bad data?” Bhalla asks. “As the culture shifts to recognize the critical and far-reaching impact of data, that sense of accountability is evolving. It’s time for credit unions to step back and look at how stronger data management can improve decisions and outcomes.”

Applying that discipline to marketing, he recommends taking a more concrete approach to setting goals for increasing returns by working with higher quality data. A good place to begin is to aim for steady gains above the baseline of returns for prior marketing activities and then to apply strategies for cleaning up member data and combining information from various sources to better target offers. Comparing results to previous campaigns will demonstrate the return on investing in better data management.

Setting benchmarks drives those efforts forward, and developing controls, a series of checks and balances to identify and correct bad data, can reduce risks and produce steady gains. CU Rise Analytics is developing predictive analytics to run simulations on marketing campaigns to compare with small-scale sample test runs, Bhalla explains. If the results are close to the simulated forecast, that confirms the credit union is using the right data, so it can launch the campaign to the full targeted audience with greater confidence. 

Apply sound logic. In implementing analytic tools, managers must be on the lookout for algorithms with built-in biases that may be hard to spot and hold the potential for discriminatory decision-making, Manus advises. “Make sure that you have enough rich, clean, quality data from diverse sources in enough volume … and then test the results to make sure you have confidence in the predictions and outcomes.

“My high school teacher always said, ‘Show me your math,’ and that same principle goes into advanced analytics,” she adds. “If you cannot tell which steps you took, you don’t know what steps you didn’t take. If you don’t know what questions you asked, you don’t know what questions you didn’t ask.”

Supplement data know-how. As credit unions develop their in-house expertise, especially for high impact data-based decision-making, they can work with experienced third parties to improve data management and analytics. For example, CU Rise Analytics has developed an attrition model that identifies in real time, based on data across systems, which members may be moving away from their credit union toward other financial service providers. Signs of disengagement include the disappearance of recurring direct deposits or bill payments, a decrease in the number of debit and/or credit card purchases, and changes in members’ shopping habits or locations.

“Any significant change in pattern could signal a shift to other financial institutions’ products,” Bhalla notes. “These models are constantly monitoring for behaviors such as sharp decreases in account balances and transaction volume. Accurate data and timely follow-up are crucial in stemming attrition.”

There’s a lot of interest in data analytics among credit unions these days, but Trellance recommends that its clients conduct a thorough review of their foundational data management capabilities before committing to develop and implement complex analytic systems. Caldera cites a Wall Street Journal article offering a caution from an IBM executive that 80% of the work in developing AI projects involves collecting and preparing data.

“Sophisticated data analytic models are very sensitive to bad data, so before investing in that level of analytics, you need to ensure that the data is sound. Any suspect data will impact the results generated by those models. Pristine data quality is important,” Caldera says. 

“A common problem is that many organizations have so much data they don’t even know where it all is,” he adds. 

To go from bad data to good, differentiate between data vision and strategy, Manus advises. The vision sets out why the credit union aims to develop data capabilities, and strategy specifies how the organization will move its people, processes and technology forward to improve governance, business maturity, change management and cultural readiness. 

“Having a budget and road map in place takes you from where you are to a higher elevation—to mature your ‘data muscle’ along that journey,” she says. Toward that end, AdvantageEdge Analytics has launched a consulting service to guide credit unions in their data strategy development and implementation. 

It’s difficult to overstate the complexity of the job ahead for many credit unions to improve the quality, depth and diversity of data in order to take advantage of sophisticated analytics, or “to use data in the rearview mirror to predict what will happen through the windshield,” as Manus puts it. Rooting out bad data and aggregating more reliable and complete information will power those efforts.   cues icon

Karen Bankston is a long-time contributor to Credit Union Management and writes about membership growth, operations, technology and governance. She is the proprietor of Precision Prose, Eugene, Oregon.

Compass Subscription