Backlash over Chime account closings highlights risks in fraud detection

The neobank Chime is coping with growing pains, including a surge in customer complaints about suddenly closed accounts.

At the end of June, the data firm Apptopia declared Chime to be the most-downloaded neobank app in the U.S., with 6.4 million new installs in the first half of 2021. The San Francisco company now has an estimated 12 million users, and it’s valued at $14.5 billion.

“For those cases where Chime did make a mistake ... we sincerely apologize,” a company spokeswoman says. Yet "we are proud of Chime’s robust anti-fraud efforts, which have returned hundreds of millions of dollars to state and federal agencies during the pandemic."

But on Tuesday, the nonprofit news organization ProPublica ran a lengthy investigative report that found many Chime customers had had their accounts closed without warning and were unable to access their money.

The case highlights a challenge every financial company struggles with: detecting fraud and suspicious activity without inadvertently red-flagging innocent customers. That task has only grown harder lately as many people’s financial behavior has been skewed by the need to conduct many activities remotely, an influx of short-term government aid and other hallmarks of the pandemic.

ProPublica told the story of Jonathan Marrero, who took his 5-year-old twins out for lunch at an Applebee’s the day after receiving a federal stimulus payment. Earlier that day, he had seen a balance of nearly $10,000 in his Chime account.

“When he went to pay, his only means of payment, a debit card issued by the hot financial technology startup Chime, was declined,” ProPublica wrote.

He then was unable to log into the Chime app. He checked his email and found a message from Chime that read, “We regret to inform you that we have made the decision to end our relationship with you at this time. Your spending account will be closed on March 18, 2021.” He had to ask his parents to pay his lunch bill and was deeply embarrassed.

There have been 920 complaints filed about Chime in the Consumer Financial Protection Bureau since April 15, 2020, according to ProPublica. Many of the complaints involve closed accounts, the publication said. Customers have also filed 4,439 complaints against Chime at the Better Business Bureau. By comparison, there have been 3,281 such complaints filed for Wells Fargo, which has a far larger customer base.

Chime told ProPublica that the closed accounts were a result of its efforts to crack down on fraud related to stimulus and unemployment payments.

A Chime spokeswoman told American Banker on Tuesday that many complaints submitted on the CFPB’s portal — and their sources — are hard to substantiate. Chime was able to prove that fraud or violations of Chime’s terms were involved in some of the instances ProPublica reported, she also said.

“For those cases where Chime did make a mistake, including the two highlighted in this story, we sincerely apologize,” the spokeswoman said. “We have made efforts to make things right with these members. We are proud of Chime’s robust anti-fraud efforts, which have returned hundreds of millions of dollars to state and federal agencies during the pandemic."

Chime is not the only neobank that’s heard from unhappy customers who can't access their money.

A number of consumers recently lashed out at Varo online for closing their accounts unexpectedly and without a clear explanation.

Customers eventually got their money back. Some received messages from Varo stating that their accounts had been closed in error.

CEO Colin Walsh said in April that the incidents resulted from suspected fraud.

“That is definitely not something that is an ongoing or recurring pattern,” he said.

Fintechs like Chime and Varo, like all financial institutions, aim to develop technology that is sophisticated enough to detect the activity of scam artists and criminals yet not so sensitive that it’s triggered by the unusual characteristics or behavior of good customers.

A recent Supreme Court case shows the danger of simplistic, rules-based fraud and anti-money-laundering software. In 2002, TransUnion introduced an Office of Foreign Assets Control Name Screen Alert. When a business opted into the service, TransUnion would conduct its ordinary credit check of the consumer, and it would also use third-party software to compare the consumer’s name against a list maintained by OFAC — a unit of the U.S. Treasury Department — of terrorists, drug traffickers and other criminals. If the consumer’s first and last name matched the first and last name of an individual on OFAC’s list, then TransUnion would place an alert on the credit report indicating that the consumer’s name was a “potential match” to a name on the OFAC list. At that time, TransUnion did not compare any data other than first and last names.

This led to a class action: 8,185 people with OFAC alerts in their credit files sued TransUnion, saying the company failed to use reasonable procedures to ensure the accuracy of their credit files. The Supreme Court ruled that only those who had actual harm done were eligible for compensation. For instance, in some cases, lenders denied credit to people because of these OFAC alerts.

Fraud and crime-detection technology have evolved since 2002, when U.S. businesses were still reeling from the World Trade Center attack of Sept. 11, 2001.

“The circumstances that led to this case were quite some time ago, and they happened at a time when everyone was still getting used to a lot of the Patriot Act requirements,” said Stephen Newman, partner in the New York law firm Stroock & Stroock & Lavan. “Across the industry, everyone's using techniques that are more refined. There've been tremendous improvements in computer technology even in the past five years.”

Machine learning has been a big part of the advances. Companies like PayPal and Synchrony, for instance, have improved their ability to detect fraud and crime through the use of such software, which learns over time to understand people’s habits: where, when and over what devices they buy things and conduct banking transactions.

Chime executives have spoken publicly about the company’s use of machine learning fraud-detection software from Simility, a company that was acquired by PayPal in 2018.

A case study posted on the Best Practice AI website states that Simility’s software helped Chime decrease fraud losses by 40%, from 18 basis points to 11.

The Chime spokeswoman declined to comment on the company’s use of Simility software.

Any AML software is subject to errors of two types: false positives and false negatives, pointed out Todd Baker, a senior fellow at the Richman Center for Business, Law & Public Policy at Columbia University and the managing principal of Broadmoor Consulting.

“Errors happen everywhere and to regular banks as well as neobanks,” Baker said in discussing Chime's issues. “In this case, the software detected fraud and the partner bank's policy was followed, which apparently required that the account be closed and in some cases that the deposit be returned to the U.S. government sender.” Chime’s bank partners are the $6.2 billion-asset Bancorp Bank in Wilmington, Delaware, and the $1.17 billion-asset Stride Bank in Enid, Oklahoma.

Chime says it has been trying to meet or exceed federal anti-fraud guidelines at a time of major economic disruption and massive unemployment benefits fraud.

"In partnership with our bank partners, Chime has continued to deploy new detection systems and adopt new rules as we’ve adapted to the increase in attempted fraud," the company spokeswoman said. "Total account closures remain very small, a fraction of a percent of our overall member base."

Chime’s real problem is customer service, Baker said.

“As the primary customer service interface for the banks, Chime needs to automate its operations as much as possible to keep its costs low,” he said. “That’s because it has an extremely limited revenue base — essentially its share of debit interchange revenue — that can’t support costly human-centered customer service. So it pushes customers to email and text-based chatbots and away from phone contact. The result? Frustration when something goes wrong. In this regard, the neobanks are worse off generally than standard banks which typically make it easier to get to a person to resolve issues.”

Many fintechs are in the same boat.

“You saw it with Robinhood when it misreported balances and when it had to halt trading in meme stocks,” Baker said. “Automation works great when things are going well, but customers get angry when their problems aren’t resolved quickly.”

For reprint and licensing requests for this article, click here.
Fintech Fraud detection Differentiated Data and Advanced Analytics
MORE FROM AMERICAN BANKER