From 2008 to 2010, as the United States plunged into the worst economic crisis on record since the Great Depression,...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
the nation witnessed 322 banks shutter their doors, according to the Federal Deposit Insurance Corporation (FDIC). Comparatively, only three FDIC-insured banks closed in 2007.
It was the kind of message no financial institution could ignore, with even some of the largest standing on shaky ground. The message was punctuated by new federal regulations, enacted with the passage of the Dodd-Frank Act.
For Zions Bancorporation, a commercial bank holding company based in Salt Lake City, Utah, the new reality has, in part, increased its focus on modeling and risk management efforts around various loan portfolios. That has required a technology upgrade to collect, store and analyze more data than it has in the past.
Before, Zions analysts could, for example, look at a pool of loans, monitor its performance over time and compare the data to other aspects of the bank’s performance.
“Historically, we’d use simulation and regression analysis tools to do that work,” said Clint Johnson, senior vice president of data warehousing, business intelligence and analytics for Zions.
But, Johnson said, the company, which holds eight commercial banks all sharing the same technology service provider, couldn’t drill down and build predictive models on individual customers, helping pinpoint who may be likely to default.
“We needed bigger tools and more horsepower to do that type of analysis,” he said.
To expand its data management environment and build its predictive analytics program, Zions purchased a data warehouse from Greenplum to replace its Oracle platform and in-database analytics software from Alpine Data Labs, a startup that had spun out of Greenplum. Greenplum was later acquired by Hopkinton, Mass.-based EMC in 2010.
New technology, old culture
As the data management needs and the number of users who wanted access to that information grew, Zions faced having to invest in additional single-database servers -- a costly endeavor, Johnson said.
“Traditional database technology meant that the data you had to work with, the number of users you had to support, the volume you had to manage were all constrained to fit the size of a single database,” he said.
Or the company could find an alternative solution. In late 2008, Zions implemented Greenplum’s massively parallel processing (MPP) data warehouse, a single system that uses multiple processors so that it could bundle together commodity servers and spread out the workload.
“The expectation in making the shift from our old Oracle technology to Greenplum is that it would enable new interactions with the data and new analysis to do things we’ve never done before,” Johnson said.
That goal has been realized in data storage alone. Johnson said before the transition, Zions was managing around 3 TB of data. Since installing the MPP data warehouse, that number has grown to nearly 12 TB, and in the next three years, it is expected to more than quadruple.
The number of users accessing that data increased from about 200 employees to close to 600.
Installing a new data warehouse platform not only meant a shift in technology, but in the culture of the institution as well.
Knowing that some could be reluctant adopters, Johnson said the organization spent time training and coaching the technical team to make sure it had the support it needed; it was also conscious the transition needed to be transparent and create minimal impact on the line of business. Most business users interact with the database using SAP’s BusinessObjects, Johnson said.
“We didn’t want them to have to [for example] rewrite any reports,” he said. “We were careful of implementation from the user perspective that everything remains the same.”
Next step: ‘Big data’
Still, the financial services institution sought to branch out to new ground -- to move into predictive analytics and data mining.
“It was one aspect we wanted to get into,” Johnson said. “These types of strategies are not new. Even to banking.”
While Johnson was reluctant to share the names of other vendors the bank investigated before investment, he provided shades of details, describing some as industry giants as well as open source practitioners.
Zions eventually selected the San Mateo, Calif.-based Alpine Data Labs, a startup that released its initial in-database predictive analytics offering called Alpine Miner in May. The software company’s original release was designed for its root company, Greenplum. That alone gave Zions incentive to invest.
“We looked at other technology providers in this space,” Johnson said, including products that didn’t have an in-database offering. “All were costly and would have required our infrastructure to be built out. … When Alpine came along, we didn’t have to make an infrastructure implementation to make it work.”
Alpine Miner runs inside the MPP database, which is capable of accessing all of the structured data contained within, according to co-founder and CEO Anderson Wong. Unlike in-memory analytics, users are able to build predictive models that take advantage of large data sets and perform scoring inside the database without having to move the data to another location.
For Zions, the new technology not only means building models on possible loan defaults, but also keeping tabs on customer behavior in the form of, for example, transactional data at ATMs and online.
“That kind of data is probably the best data to mine when looking for customer behavior,” Johnson said. “To understand what the customer is doing, we needed to have the ability to process, monitor and evaluate what’s really happening within that transactional data.”
Johnson said this is particularly useful when it comes to customer turnover for the retail banking industry, which can see fluctuations of 18% to 20% per year.
“We would prefer to hold on to those customers,” he said. Changes in behavior may signal customers are ready to pack their bags, and, by recognizing the signs, the bank may have a fighting chance to address the customers’ concerns before it’s too late.
“We feel this is worthwhile and cheaper than letting customers walk out the door,” he said, adding that acquiring new customers also comes at a price.
While Alpine Data Labs bills its in-database analytics offering as “intuitive” and “straightforward,” utilizing user-friendly visualization tools for “all levels of the organization,” only a few Zions employees use the offering.
“It’s a small set of data scientists who work with Alpine and Greenplum … because [the work] requires strong database skills and strong analytics skills,” Johnson said.
Last week, in the hopes of taking on predictive analytics providers such as SAS and IBM’s SPSS, Alpine Data Labs announced a 2.0 release of the Alpine Miner product, expanding its offering to Oracle’s Exadata and the open source PostgreSQL database.
Wong said Alpine Data Labs is also working closely with Hadoop to develop a version of the Alpine Miner capable of taking on semi-structured and unstructured data. The company hopes to release that product in just over 18 months.