Silicon Valley Wants to Use Algorithms for Debt Collection

Startups hope to disrupt the robocalls and boiler rooms by harnessing data to make collection both gentler and more efficient.
Silicon Valley Wants to Use Algorithms for Debt Collection
Casey Chin

Consumer debt, credit card debt, and personal loan debt are at all-time highs. Meanwhile, investors who purchase debt for cents on the dollar and then try to collect the whole amount, and the collection agencies they hire, are getting increasingly aggressive. One in four consumers contacted by debt collectors feels threatened, and most consumers say the calls persist even after requests to stop, according to a 2017 study by the Consumer Financial Protection Bureau.

To borrowers who owe money, it’s a living nightmare. To Ohad Samet, cofounder and CEO of TrueAccord, a San Francisco debt-collection startup that has raised nearly $30 million, it’s a software problem.

“We believe that we can use technology to radically change the user experience and really help people with their day-to-day finances,” he says.

Instead of robocalls that go unanswered, letters lost in a pile of mail, and pushy collection agents who work on commission, TrueAccord contacts people through email, text, and the occasional Facebook ad, nudging them to check their inbox for an email from TrueAccord. Customers can adjust repayment plans online, changing the amount week to week or canceling a payment with no fee.

The company uses machine learning to analyze data collected from behavior on its website and other information shared voluntarily. TrueAccord says it does not buy any personal, financial, or demographic data, including credit scores, does not use affinity data, and does not “creep crawl the web.” But it does know how much a debtor owes, to whom, and how far behind the person is on the payments. Over time, the company believes, this data will help it predict preferences, like whether customers prefer text versus email, days and times to send messages, and even tone of voice, such as empathetic, friendly, or inspirational, but never aggressive.

TrueAccord is part of a wave of tech startups that claim they can increase the amount collected on debts and help consumers at the same time by using technology to personalize the process. Y Combinator’s demo day in August featured a pitch from the incubator’s second debt-collection startup, Prodigal Technologies, which says its software can make it easy for borrowers to explain their financial situation by uploading unemployment or insurance forms. Y Combinator, which is simultaneously developing a large-scale basic-income study, also backed a medical-debt company called Collectly, which has raised nearly $2 million. The companies tout flexible repayment plans, but lenders restrict what they can offer. Prodigal’s CEO Shantanu Gangal says his company works with a lender’s risk, compliance, and operations team to generate options modeled on the lender’s past data. “To the extent possible, we will come up with another borrower-friendly alternative,” Gangal says.

China has also seen a burst of debt-collection startups, including Ziyitong, which uses artificial intelligence to scrape the internet for information on borrowers and their friends, and Yigou, which provides collection agents with geolocation data on some borrowers.

In comparison, the American variety sounds more like ecommerce and less like surveillance, which is part of their pitch. Treat debtors more like online customers early in the process and you can save them from less-friendly players down the line.

Venture capitalist Hunter Walk, who invested in TrueAccord in 2013, said his firm, Homebrew, was drawn to the idea of keeping debt out of the “hands of increasingly aggressive, shady, and sketchy collection agents.” He said the firm first mulled whether it wanted to wade into debt collection, drawing a comparison to ecigarette companies. “Even if vaping is healthier than cigarettes, I'm not sure I'd want to be in the tobacco business, PERIOD,” Walk said via email. “But our research and conversations with Ohad suggested that TrueAccord wasn't simply a ‘less evil’ product but one that at scale could help consumers settle their debts, improve their credit, and remove stress created by their previous financial decisions.”

Ira Rheingold, executive director of the National Association of Consumer Advocates, is skeptical. “A kinder, gentler debt collector? I’m not sure I’ve seen the beast,” Rheingold says. No matter how you slice it, “they’re simply completing against other creditors to get your money quicker and faster.”

Software may improve efficiency, but it doesn’t address the underlying reason people fall behind on their bills. “They’re not paying their debt, because they don’t have the money,” and that won’t change without access to more income or job opportunities, Rheingold says.

Venture-backed companies are not your friend or your counselor. They are trying to get a slice of the pie, and the information they collect about you “just makes them a better salesperson” by refining the script for collection agents, Rheingold says.

Debt collection is just the latest sector of consumer finance that tech startups have pledged to reinvent by collecting more data and profiling consumers, a practice critics say has already led to digital redlining, where consumers are offered different deals based on how an algorithm views them. First came student loans, then payday lending, subprime loans, and credit scoring. Fintech companies like Avant, Prosper, and SoFi are now responsible for a third of the booming personal loan market, according to a recent study from TransUnion.

The plunge into finance also brings Silicon Valley, which likes to operate free of government oversight, into a highly regulated industry. That has already caused problems for some startups. In 2017, Earnest, a student loan startup, was bought by Navient, formerly part of Sallie Mae, and one of the nation’s largest student loan companies. The CFPB is suing Navient for allegedly cheating borrowers. LendUp, a subprime credit-card startup also backed by Y Combinator, paid the CFPB a $6.3 million settlement in 2016 for misleading consumers.

The debt collectors are covered by the Fair Debt Collection Practices Act, which is in need of an overhaul; the law was written before email, for example. The CFPB has been promising to write new rules around debt collection for years, but the process is now delayed until March 2019.

Consumers are not waiting around. Since 2015 there have been 28 lawsuits filed against TrueAccord in federal court, seeking damages for violations of the FDCPA. At least 10 are pending. One claims that TrueAccord allegedly caused a man in Florida undue distress by sending debt collection notices to his mom’s email address, allegedly misled a woman in New Mexico by offering three installment plans that each came out to more than what she owed, and allegedly hounded a man in Michigan with 37 emails in five months, including some that mentioned “our previous conversation,” even though none took place. Consumers have also filed 94 complaints about TrueAccord with the CFPB since 2014.

Sabet says he takes the complaints and lawsuits seriously and can’t comment on individual cases, but says the problems are “extremely rare,” considering that TrueAccord has communicated with nearly 4 million consumers. “We've been fortunate to see a fraction of these issues compared to traditional players,” he says.

Collectly, another Y Combinator–backed company, refrains from collecting more data because of compliance concerns, CEO Levon Brutyan says. The company collects medical bills for smaller practices, where the median balance is $70 to $80, he says. It uses the patient’s zip code to estimate the person’s financial condition, and the amount of the debt to text or email a payment option, for instance using a Pay Now button or a Subscribe to Plan, depending on the amount.

Brutyan doesn’t think of Collectly as a debt-collection company. “We’re patient-responsibility collection software, not a collections agency,” he says.


More Great WIRED Stories