While banks talk about developing a 360-degree view of their customers, they often can’t comprehensively identify even a single individual’s relationship with the bank, according to Pamela Pecs Cytron, founder and CEO of Pendo Systems. The information may be held in different systems; it may be stored under variations in a person’s name, and given the inevitable errors in manual data entry it might simply be spelled incorrectly or entered in the wrong field. In addition, data captured today may not even have the right fields to be stored in the legacy systems.
A top tier global bank recently began using the Pendo Data Platform (PDP) to find its credit exposure to selected group of ultra high net worth clients. In less than 24 hours, the PDP aggregated five disparate systems (loans, deposits, credit cards, collateral, and investments) never before connected by the bank, and conducted a search and analysis across all five.
Cytron calls the hidden information “dark data,” but it isn’t really dark; it’s just not necessarily where anyone is looking. Data analysts typically starts the process of aggregation by looking at column labels. If they are looking for a trader, they will look in the column labeled trade or trader. Or if they are looking for a name they might look for “Tom” but miss “Thomas”. The Pendo Data Platform looks at structured and unstructured data from systems, tables, PDFs, and even emails to generate a consolidated view that is then available for indexing, searching, and analysis.
“My first account at my bank was Pamela Pecs, then Pamela Pecs-Cytron with a hyphen and now Pecs Cytron without the hyphen,” said Pendo’s CEO. “No wonder systems get confused and lose people. And that’s just my last name. In one system I might be Pammy, Pamela, or Pam. Just looking for a name could miss me completely because they didn’t look at me in the context of who I am, but in the context of the table they expect me to be in, and that’s in just one system.”
Within a single bank, an individual might be in multiple systems such as mortgages, credit cards, brokerage, or a business name account. Data clerks are trained to look at what they are looking for; they don’t have the capacity to look everywhere for any potentially useful identity hint. The Pendo Data Platform acts like a high-speed robot that does look everywhere. It goes up and down every column and across each row, looking for useful information from names, addresses, and even zip codes – using data lineage and linking to produce valuable data for easy review and verification.
“A big customer might have multiple relationships in both the personal and commercial sides of a bank,” explained Cytron, “and those might not be aggregated in total. We have identified customers’ global exposure, even finding some through common lawyers and phone numbers.”
The PDP discovered a previously unknown very large exposure to a single individual.
The Pendo platform doesn’t work just with names and addresses; it can address other data problems. In early 2016, Pendo helped a major bank solve their CCAR (Comprehensive Capital Analysis and Review — aka stress test) issue related to loan linkage, customers with current mortgages who had defaulted on or modified previous loans. After 100 people had worked on the problem for two years, they had found 250 loans at risk. Within two months the Pendo system found 1,700 more high risk mortgages.
“Banks are reserving billions more than they need to because they can’t find the relevant data in their systems to show exactly what their risks are, and that impacts shareholder value,” Cytron said.
The Pendo Data Platform works quickly, she added, so it can produce results for regulators or board members. The technical staffs are building data lakes, which may bring data from multiple systems into one place. But along the way the correlation of the data is not being improved. It’s just another data warehouse.
