Establishing a Credit Data Mart: A 90-Day Sprint to Value

By Charles F. Clow; Brad P. Saegesser, CGMA; and Christopher J. Sifter
| 10/27/2017
Running along road


support-img-lp-perf-18101-018e
In today’s data-driven bank operations, having direct and consistent access to secure and reliable data is essential for a broad range of credit functions.


Rather than maintaining a number of disconnected and often duplicative data sources, best-in-class banking organizations are finding there are significant advantages to establishing a centralized credit data mart. Properly designed, structured, and governed, a central credit data mart can serve a variety of data needs. These include regulatory compliance and reporting requirements as well as specific strategic, operational, and customer service functions.

Although establishing a credit data mart might seem like an overwhelming assignment, there are ways to overcome the challenge and begin making progress. One key to success in large, complex projects is to achieve some early wins and establish clear, visible examples of how the effort is adding genuine value. Such a sprint to value can help overcome inertia and maintain momentum in the pursuit of improved data capabilities.

Credit Data Challenges: The Shortcomings of the Traditional Approach

Historically, banking organizations have approached data issues from a narrow perspective, focusing only on the specific data that was needed for a particular task at hand. Finding, acquiring, and validating that data typically required an exhaustive search through a variety of disconnected sources, both internal and external. These steps were then followed by the build-out of a solution that achieved one specific analytic or reporting objective.

For example, many data projects over the past few years have been driven by banks’ efforts to comply with the Dodd-Frank Wall Street Reform and Consumer Protection Act, which requires financial services companies with more than $10 billion in total consolidated assets to conduct annual stress tests (DFAST). In many cases, these projects involved extensive searches to locate and access the specific pieces of information that would be needed for DFAST reporting – information that often was housed in a variety of sources including ad hoc spreadsheets and paper documents. Eventually each bank’s relevant data would be sourced and accessed via a custom-built data mart that housed all the specific data needed to comply with DFAST submission requirements.

While the DFAST data project was underway, however, other teams began considering what data would be needed to apply the Financial Accounting Standards Board’s (FASB) new current expected credit loss (CECL) model to their credit impairment and allowance for loan and lease losses (ALLL) calculations. Eventually, this data would be located and linked to a separate, dedicated CECL/ALLL data center.

In addition to being narrowly focused and constrained, such an ad hoc approach also is inherently inefficient on a larger scale because the large number of separate data-driven processes typically draw on many of the same sources. But inefficiency is only one drawback to this traditional approach to managing bank data.

Because there is no common ownership or responsibility for the accuracy, completeness, and reliability of the data, the ad hoc approach raises significant risk that the various data sets will be inconsistent or in conflict with each other – a situation that could lead to serious regulatory and operational issues. This approach also is a maintenance nightmare since each of the many data sources is feeding multiple data marts, which in turn are operating with varying levels of control and access.

A Better Approach: The Central Credit Data Mart

As banks recognize the shortcomings of conventional data management approaches, the advantages of developing a central credit data mart become more apparent. Rather than maintaining separate data systems and sources for each credit-related process, a central credit data mart enables each data user to draw on the same consistent, reliable data sources.

Central Credit Data Mart Concept DataGovernance2

Source: Crowe analysis

Such an approach requires – but also reinforces – effective data governance practices. The objective of data governance is to see that data is treated as a business-critical organizational asset. The desired result would be that data consumers and business stakeholders could:

  • Trust that the data collected and made available is accurate, without question or hesitation, and that its flow is controlled and audited
  • Have consistent access to the information needed for their business priorities and commitments
  • Uniformly rely on the same source of data, rather than separate versions

Effective data governance also means there is organizationwide commitment and funding to maintain data quality and availability across the bank. In the longer term, it also clarifies how future needs for new data or information will be addressed, with a consistent process and reliable technology.

Getting From Here to There: A Sprint to Value

The first major challenge most banks face when developing a central credit data mart is figuring out how and where to begin. Addressing this challenge often requires an agile approach that counteracts the natural tendency to seek perfection at the outset and the fear of launching an effort until perfection can be guaranteed.

Another important factor in successfully launching a credit data mart project is the ability to achieve some early successes – a sprint to value in which the potential benefits of the project can be measured and documented quickly. Several data-intensive initiatives in particular often provide opportunities to do this:

  • Stress Testing and Capital Planning. DFAST forecasting and capital planning processes both impose substantial data requirements, many of which rely on the same or closely related sources. Both processes require the use of common external data such as call report data, as well as internal bank asset and liability management data. Much of the data needed to produce the required nine-quarter DFAST forecast also is used to perform longer-term capital planning. By establishing a central data mart, banks can share not only the source data for these two processes, but also many of the related calculations and estimates they have in common.
  • CECL Planning and Preparation. One of the features of the transition to a CECL model for calculating credit impairment is that banks have the flexibility to estimate lifetime losses in several different ways. One consequence of this feature is that banks can begin preparing for CECL by first understanding what data they have and then building models that incorporate that data. In addition to pinpointing and sourcing the available data, a central credit data mart also can help support risk segmentation and risk identification analytics.
  • Accounting Standards Update (ASU) No. 2016-01 Loan Valuation. This new FASB standard spells out how banks that are public business entities must disclose the fair value of their loan portfolios and the valuation methodology they use to prepare this disclosure. Because the new methodology uses an “exit price” approach rather than an “entrance price” approach, banks will need to apply a discounted cash flow model to calculate liquidity, risk, and other factors. A central credit data mart can provide users with clean, reliable data drawn directly from the bank’s core systems, which can then be exported directly into the loan valuation model.

By applying the credit data mart concept to high-priority initiatives such as those listed here, banks can build momentum and early recognition of the value that a data mart can offer. In a sense, however, achieving early success often requires taking a counterintuitive approach. Rather than setting out to meet the specific data needs of any particular effort (such as DFAST reporting or CECL compliance), the development of a credit data mart begins by identifying credit data sources and accumulating all the available data first. In other words, instead of starting with an end goal in mind, the effort begins by identifying the current data situation, and then builds from there.

Another important principle to bear in mind is that the initial launch of the data mart might not be fully comprehensive. Some data elements probably will be missing at launch, but new data fields can be added incrementally over time in a series of phased projects. The key is to start with the available data and the infrastructure that is already in place, and focus on gaining value quickly.

Contact us

people
Charles Clow
people
Brad Saegesser
Chris Sifter
Christopher Sifter
Principal