A New York judge’s recent ruling that lenders to financially ailing cosmetics giant Revlon can keep US$500 million in payments they mistakenly received from Citibank is prompting operations managers of syndicate loan desks at rival banks to take a closer look at their technology, internal procedures and any offshore work.
Citibank is considered an anomaly among syndicate lending administrative agents in using an application– Oracle’s Flexcube — that relies on manual intervention as well as a third-party offshore business operations firm Wipro in India. The bank planned to be switching over to a more automated system — Finastra’s Loan IQ– but had not done so by the time Citibank made a $900 million payment of instead of only a US$7.8 million interest payment in August 2020. The amount paid represented the full amount of the loan made to Revlon in 2016 plus interest. Operations managers at other banks who claim to have “well-tuned” syndicated loan operations tell FinOps Report that they belatedly realize the risk of error is cost-prohibitive, because the law might not be on the side of the loan administrator in case a payment is mistakenly wired to a lender. No one is willing to acknowledge whether his or her bank has made any mistakes in wire payments, but from what they tell FinOps Report they presumed that the lenders — or investors– would always return the funds as that was standard industry practice. That’s not what happened with Citibank. Some of the lenders did return the funds they received while others — including Brigade Capital Management, HPS Investment Partners, and Symphony Asset Management — refused and New York District Court Judge Jesse Furman ruled in their favor after Citibank sued them and others. Of the US$900 million the largest amounts were paid to Brigade Capital at US $175 million; HPS at US$133.5 million; and Symphony at US $109.67 million. The remaining seven fund managers who declined to return the funds kept US$108.7 million in total. Citibank says the entire overpayment was made with its own money.
“We are now testing all of our systems, will speak with vendors, and review our procedures,” says one syndicate operations manager representing the consensus of ten executives contacted by FinOps Report “We tell clients it won’t happen to us, but we can’t be sure unless we do some more trial runs of the technology.” Apparently, there is a concern that even the most automated system might not catch human errors in data entry which is what occurred with Citibank. The operations managers tell FinOps Report their banks will also be adding more layers of approval in the event of a large payment or a payment that is not made on schedule. Their banks will also be implementing additional training for junior and senior level professionals to be up to speed on their workflow process and technology. None of those questioned by FinOps Report — all using Finastra’s Loan IQ– wanted to disclose whether they relied on offshore staff for their syndicated loan operations. (Goldman Sachs uses home-grown technology). However, all insisted that if they did it they would likely also be reviewing those procedures as well. Although some syndicate loan administrators — such as JP Morgan Chase and Bank of America– do have offshore operations centers — they rely on their own staff to do the work.
Some of the operations managers at rival banks contacted by FinOps Report to discuss Citibank’s mistake criticized Citibank for relying on an external offshore operations provider — Wipro–while others blamed the “convoluted application” Citibank uses and yet others blamed poor procedures. Based on court documentation on what went wrong a combination of human error and legacy technology appears to have been the case. Once the data was input incorrectly, there was a good chance the wrong payment would have been made based on how the Flexcube system worked. The Flexcube system, insist some syndicated loan operations managers, is not designed for the syndicated loans market. The software was initially developed by a Citibank spinoff I-Flex Solutions, which agreed to replace Citibank’s legacy system and in 2005 Oracle bought Citibank’s 41 percent stake in the company.
Citibank’s spokesperson would not say how long it has been using the Flexcube system– over a decade by industry estimates — or when it would move over to another system, or acknowledge the identity of the new platform supplier. She would only say that in 2019 Citibank decided to “upgrade” its loans processing system. Finastra declined to confirm Citibank’s pending use of Loan IQ or provide any other comment for this article.
As a syndicated loans administrative agent, Citibank is one of a handful of firms hired by borrowers — typically large corporations — to calculate the principal and interest payments due to all the lenders and make them at the appropriate times. If Citi hopes to recoup the US$500 million it mistakenly paid out it must now appeal Judge Furman’s decision that its mistake falls under New York’s exception to the rule of returning funds received incorrectly. The fund management firms which refused to return the funds paid last August, says the judge, can keep the money because they had no reason to believe Citibank made a mistake; they just can’t spend the money until Citibank’s appeal is over. Judge Furman apparently bought the explanations of officials at lender organizations that they thought Revlon was paying down its loan early. After all, the amounts they received were the exact amounts “to the penny” of what Citibank should have sent them, albeit in 2023. Naturally, Citibank disagrees with Judge Furman’s ruling. “We believe the law is on our side and we will recover the outstanding funds,” says the bank’s spokesperson. The Loan Syndications & Trading Association, the New York-trade group for the US syndicated loans market, has publicly called Judge Furman’s decision potentially detrimental to the industry.
According to documentation presented during its December 2020 trial, Citibank initially blamed the Flexcube application for its mistake but subsequently acknowledged it was human error. Six eyes — or three different individuals — were assigned to make certain that payments to lenders were accurate. Of the three employees at fault, two worked for Wipro and the third was a supervisor at Citi in Delaware who had the final say. The fact that the work was done remotely during the COVID-19 pandemic didn’t play a role in the error. None of the three employees relied on Citi’s manual for using the Flexcube system which indicated there were multiple data fields that needed to be filled out and adjacent boxes on the Flexcube user interface that were supposed to be manually checked off to ensure that only US$7.8 million would be paid out and the other millions of dollars would remain in an internal or “wash account” at Citibank.
According to the bank’s user manual any payment entered into the Flexcube platform is released “unless the marker suppresses the default option.” As the first data entry specialist at Wipro — otherwise known as the marker– Santosh Ravi– filled in the data field for principal, then included the number of Citibank’s internal or wash account and then clicked on the overwrite default settlement box next to the data field for principal. However, that step alone would not have prevented the entire principal amount from being prepaid. Two other fields called “front” and “funds” should also have been filled out and the overwrite default settlement boxes next to those fields should also been checked off. Arokia Raj, the second employee at Wipro — otherwise known as the checker — approved Ravi’s work, because he made the same assumption that filing in the data field for principal with a wash account number and checking the overwrite default settlement box would be sufficient. Vincent Fratta, the syndicated loan operations supervisor from Citibank in Delaware, also approved the data entry work reaching the same conclusion. “Raj, Ravi and Fratta believed incorrectly that the default instructions were properly overriden and principal payment would go to a wash account,” acknowledges the court documentation. Fratta joined Citibank in 2006 and in 2012 he was named to lead the team of six Delaware-based Citibank employees and nine Wipro employees in India dedicated to Citibank. It is unclear why the error in the Revlon loan payment would be the first time Citibank encountered such a problem other than the fact it might have been the complex nature of the transaction in that an interest payment had to be made and funds transferred to a wash account simultaneously.
Based on how the Flexcube system worked. There was no way for Raj, Ravi, or Fratta to have known the wrong payment would be sent until it was too late. Once Raj received the authorization from Fratta to make the payment, upon pushing the pay button a stop warning did appear on Flexcube’s user interface screen. However, that alert never clearly indicated that the principal payment was being wired out of the bank or any information on the amount of payment that would be made. Instead, the screen said: “Account used is wire account and funds will be sent out of the bank. Do you wish to continue?” Raj likely believed that only US$7.8 million would be wired out and not the remaining principal amount. Once the entire payment was made, Flexcube’s screen said the following “US$7.8 million in interest and US$894 million in principal paid.” At that point, according to the court documentation, Ravi incorrectly thought the principal payment had been sent to Citibank’s wash account and only the interest payment was made to lenders. Whether he should have known otherwise is debatable. Some syndicate operation managers at rival banks say the phrase “principal paid” should have alerted Ravi to the fact the total principal amount payment had been erroneously wired to lenders while others say that based on his lack of understanding of how the Flexcube system worked during the time of data entry, he would not have understood the phrasing about payment. What all of the syndicated loan specialists acknowledge is that had the Flexcube screen actually indicated the full amount of the payment that would have been wired to lenders when it was observed by Raj, the blunder might have been avoided. It is unclear why that didn’t happen– was there a flaw in the Flexcube system that no one could have been anticipated is the question no one seems willing to answer. All syndicate operations managers using Finastra’s Loan IQ system at rival banks tell FinOps Report that Finastra’s application would have had a summary screen showing the full amount of the payment that would have been made– and to whom– and even an alert that another approval was needed. (That explanation could not be confirmed with Finastra).
It was not until the day after the payment was erroneously made that Citibank’s Fratta caught wind of the blunder when the bank reconciled its books. At that point the damage was already done. Fratta might have thought all the lenders would return the funds once Citigroup notified them of a mistake, but there was one additional factor that made it unlikely– bad blood between the lenders and Revlon. Concerned about Revlon’s restructuring of the loan. some lenders had sued Revlon demanding earlier repayment and had reason to believe they would not receive the full amount of the loan in the future. Hence, they were only too happy to receive the money early and persuade Judge Furman they thought it was a legit payment. Those same lenders had also demanded Citibank resign as the administrative agent and Citibank had reportedly agreed. Had Citibank done so by August 2020, making Revlon’s interest payment to lenders that month would have been another bank’s problem.
At the very least Citibank’s blunder will bring highlight two inefficiencies in the syndicated loans market that other banks will want to review. The first is that nothing good ever happens when manual work is involved regardless of any checks and balances; the second is that relying on third party service providers may not be a good idea. Should Citibank lose it appeal it will have to negotiate with Wipro as to how much Wipro will cover in uninsured liabilities as it is likely the full amount would be covered by Wipro’s insurance policy. Wipro did not respond to e-mailed requests for comment. Full in-house automation is the best remedy, but that can only be achieved, some syndicated loan operations experts tell FinOps Report, through distributed ledger technology, Using smart contracts coded in stone to determine who gets paid, how much and when, would never have allowed an unscheduled payment of the principal amount of the loan could never have been made. Hopefully, the next phase of the syndicated loan market’s maturation will come sooner than later.