Latest / Follow the Lies: Fraud & Deceit Uncovered / Episode 17: Four Million Fake Students — How Charlie Javice Sold Frank to JPMorgan
Transcript
- riverside edit kevin kevin barrows's stu (15): It's 2021. Squid Games is everywhere. The Friends cast is back on television, and JP Morgan Chase wants a pipeline of young customers. A founder named Charlie Javiz tells the bank her company, Frank, already has 4.25 million of them. Chase pays $175 million. Months later, the bank sends a marketing email to a sample of those names. Most of the messages never arrive. Almost no one opens what does. Today we follow the lives from a four million user pitch and One of the most expensive customer lists a US bank has ever bought. Welcome to Follow the Lies for and to seat uncover. I'm Kevin Barrows. After almost three decades investigating financial crime in the United States, I had a front row seat to how deception works when numbers are treated as fact. As a former FBI special agent, I led major white collar cases involving securities fraud, bank fraud, and large scale financial manipulation. Today as a private investigator, I still help companies, institutions, and people trying to figure out what they're actually looking at. This fraud did not invent a company that didn't exist. It invented a customer count that the buyer was willing to pay for. Frank was a real company. Students actually used it. But what Chase paid for was not the software. It was the four million users. That number was not a customer list. It was a model built to look like one. If you're enjoying these episodes, I'd really appreciate it if you'd hit subscribe, leave a rating review, and tap the bell so you never miss an episode. Charlie Javiz graduated from the University of Pennsylvania's Wharton School in 2013 and built Frank to help students complete the free application for federal student aid. were FAFSA, the federal form used to apply for college aid. The legal name was TAPD Inc., doing business as Frank. The government already offered the form for free, but Frank's pitch was to streamline the process, reduce abandoned applications, and lock in a relationship with a student at the moment they first needed a bank. She said the name was a promise. Frank, she said, just meant honest. Money quickly followed the story she was telling. 2017 an Israeli firm, Aleph, Put in the first institutional check. Later that year, Mark Rowan of Apollo Global Management led the Series A, which is the first round of venture capital funding. More money followed from education and venture investors. Forbes put Javis on the 30 under 30 list. She wrote about student aid in the New York Times and appeared on CNBC. By 2019, she was pitching bankers from Lion Tree, a firm that advises company on sales and public offerings. In 2021, Frank hired Lion Tree to find a buyer. By late 2020, the user account was already being built for a sale, not for the product. What a buyer would actually be purchasing was not turbo tax for financial aid, it was a list. Banks had spent years failing to own the first checking account of a college freshman. The founder who claimed millions of those names already sat in her system was offering a serious shortcut. Jamie Diamond, the CEO of JP Morgan Chase, Had told investors the bank needed to grow more aggressively. FinTech and Big Tech are already competing for the same customers. An existing franc investor sent Javise's profile inside the bank, but early talks went nowhere. Lion Tree reopened them and kept the sale moving. In July, Diamond sat with Javise. In later court filing, she said he told colleagues to get the deal done before the school year started. When the bank announced the purchase that September, Public statement said Frank had reached 5 million student mark at more than 6,000 campuses. Javiz told JP Morgan that Frank had 4.25 million users, which she defined as people who had created an account and given four fields: first name, last name, email, and phone. A visitor is not a user. A click is not an account. Trail evidence put the real figure in about 300,000. They sold the same number to more than one bank. In June 2021, they told another acquirer, identified in court as Capital One, that Frank had 4.25 million users using the same four-field definition. Capital One walked away. Chase did not. They also put the inflated number in writing where a buyer would look first. A data room spreadsheet titled User Breakdown claimed more than 4.265 million students had started a FAFSA through Frank, and more than 2.1 million had finished one. Trial evidence put In July, Javis was still selling growth. She emailed Leslie Wims Mars, JP Morgan's head of corporate development, that a single marketing blast had created 200,000 new accounts, and that millions of people hit the site every month. Frank's entire history did not support those figures. Then the bank did the one thing an inflated metric deal cannot survive if it's honest. It made data validation a condition of closing. That same day, August 1st, 2021, Javiz told Oliver Omar, her chief growth officer, they needed to talk about a data pull and a strategy, and he should block off the day. They called each other 19 times. That afternoon she asked Patrick Vover, Frank's director of engineering, to generate 4.25 million fake customers with the same statistical shape as the real ones. Vover asked if it was legal, and she answered in substance that she did not intend to end up in an orange jumpsuit. Wisely, he said no. She then texted a college friend, Adam Kapelner, an associate professor of mathematics at Queens College. She said she was in an urgent pinch and asked if he still did consulting work. She did not tell him a bank was about the tested number, that was not true. Kapelner was not a stranger she found online. They had known each other since Wharton. He quoted three hundred dollars an hour. She offered six hundred. When he asked what the project was for, she would not say. In later government filings, The cover story she used was privacy. The synthetic records were needed to protect real customers. He was not told the records were the customers. She sent the file a fewer than 300,000 real users and told him she needed him to turn it into a file with over 4.265 million customer records. After the call, she texted Omar. I found my genius. He says it'll take him an hour. Kapelner worked through the night. He created look-alike records, synthetic people, synthetic fields, and By recombining the real seed file. At trial, he walked the jury through one of those rows: name, address, phone, email. Then he said the sentence the government wanted on the record. That person does not exist. On August 5th, at Javiz's direction, that file went to Axiom, the vendor JP Morgan had chosen. She instructed Axiom not to give JP Morgan any additional background on the file. After the report went to the bank, she asked Axiom to delete the data. Axim reported what it had been handed four million two hundred sixty five thousand eighty five rows. The bank treated the count as confirmation. She asked Kaplaner to keep the invoice vague and paid him about eighteen thousand dollars instead of the thirteen thousand three hundred dollars on the original bill. At the same time Kaplner was creating his spreadsheet, Amar went to a data broker, ASL Marketing, and purchased four point five million student records, two point eight million with email addresses and and one point seven million without, at two point three cents a name. The cost was about one hundred and five thousand dollars. He told the broker they needed it that day. If the purchase could not close immediately they would take the business elsewhere. Before the paperwork was signed, Javiz had the quantities removed from the invoice. The rented file did not have every field she had already defined as a user. They brought more names on the open market to fill the gaps. On August fifth, Amar texted her that she would have four point five million users that day. The purchase had just closed. She wrote back perfect. Love you. Those names were not Frank users. The next day she told Kaplaner the ASL file was Frank's user database. They knew Chase would eventually try to contact these people. A rented list with some working inboxes is what you buy when you expect the buyer to test whether the contacts are real. On august eighth, the merger agreement was signed. In September, the bank announced that it had paid one hundred seventy five million dollars. Javiz joined as a managing director. She received more than twenty one million for her equity. A twenty million retention bonus was on the table if she stayed. It was never paid. Amar joined the bank with her and received more than seven million dollars. If you do the bank's own math from that week, one hundred seventy five million divided by four point two five million claimed users is about forty one dollars a student. Against three hundred thousand real accounts, it's five hundred eighty three dollars per student. A very different deal. The break in the case did not come from a whistleblower. It came from an email blast. After the deal closed, Chase Marketing asked for the customer list so it could market student accounts. What came back was not Frank's real user file. It was the rented list and the fabricated file. In july two thousand twenty two, the bank sampled about four hundred thousand email addresses. Only about twenty eight percent of the messages were even delivered. Chase was used to delivery rates in the high nineties on comparable campaigns. Open rates were almost non existent. About one point one percent of what was sent was even opened. The campaign produced about ten new accounts, and none were students. Inside the bank, the test was called disastrous. The bank started comparing what it had actually received with what it had been told. The names did not look like students in the middle of completing a FAFSA. The emails did not look like a four year customer base. The row count Axiom had certified was simply a count of a file that the seller had uploaded days earlier. The line had not stopped at closing. In january twenty twenty two, the company was still sending the bank files labeled as franc users. One marketing list had about a million rows. It did not contain a single franc account. It was broker data under a franc file name. Around the same time the internal story had grown again. Five point six million users, with millions written off as no longer marketable because they had graduated. In November 2022, the bank fired Javiz and in December it sued her in Delaware. In January 2023, it shut down Frank's site and pulled the press release that had advertised five million students in 6,000 campuses. On the next earnings call, a Wells Fargo analyst pressed Diamond on who was accountable. Diamond did not answer the question and simply called the acquisition a huge mistake. In april twenty twenty three, federal prosecutors in Manhattan and the SEC filed on the same day. Criminal charges and a civil fraud case hit the news cycle together. A grand jury later returned a superseding indictment against Javis and Amar, conspiracy, wire fraud, bank fraud, and securities fraud. The government's theory was narrow. Prosecutors did not have to prove Frank never helped the student. They had to prove the 4.25 million users was a number the defendants knew was false when they used it to close a $175 million sale, and that they built and bought data to get through the bank's scrutiny. The defense called it buyer's remorse. They said Chase rushed the deal, never pinned down what user meant, and bought the story it wanted to hear, then blamed the sellers when diligence failed. The jury had to decide what that meaning and that file really were. Was user a sloppy word the bank never defined or a number the seller certified and sold? The trial began in february 2025 and ran for about six weeks. Volver described the request he refused. Kapelner walked the jury through the file he built, the invoice he was asked to clean up, and a row for a woman named Catherine Gordy. Then he said that person does not exist. He said he asked the purpose of the project, and he was told she could not talk about it. Broker witnesses put the 105,000 student lists purchased on the record, along with the rush to get the names that day. A second vendor filled the fields the first file did not have. Neither Javis nor Amar testified. On March 28, 2025, after hours, not days, of deliberation, the jury convicted both of them on all four counts. Sentencing took place on September 29, 2025. Prosecutors asked for 12 years. Javiz wrote the judge that she accepted the verdict and took full responsibility. After character letters, he sentenced her to eighty five months and three years of supervised release. He called her a good person who had done a bad thing. He entered a forfeiture judgment of twenty two point four million dollars against her, and a separate restitution order of two hundred eighty seven point five million, joint in several with Amar. The restitution figure included what the bank paid for Frank, costs after the close, and And more than a hundred million dollars in defense fees JP Morgan had been required to advance. The judge criticized the bank's diligence in open court and said he was punishing her conduct, not the buyer's mistakes. He recommended she serve the sentence at Danbury. Amar was sentenced on november fifth, two thousand twenty five to sixty eight months. The judge said he was not the instigator, but was intimately involved. Javise has remained free on bail pending appeal. In march two thousand twenty six, Her new trial motion over alleged conflicts involving law clerks was denied. The court left the verdict in place. Now let's go inside the investigation, where I pull back the curtain on how a case like this actually gets made. I've worked enough financial fraud cases to know the hard part is rarely proven the company exists. The case turns on intent. What she told the bank it was buying, and whether she knew that number was false the moment the bank asked for proof. The engineer who refused the fake file is how you prove they understood the legal risk. He asked if it was legal. She talked about an orange jumpsuit. That's not a misunderstanding. That's knowledge. Then you look at who was still on the emails after she asked the engineer to manufacture the file. Amir did not found the company. He was still on text when the broker list closed. You'll have four point five million users today, is why this was charged as a conspiracy. The jury decided he was not just an employee following orders. Cooperation was simpler here than in some of the cases we've covered. The engineer refused to build the fake file. The professor built the model without being told the user count was a lie. Those two people could testify without cutting a deal to stay out of prison. That's a significant advantage for the government. It still leaves an uncomfortable fact on the table, and the judge said it out loud. The buyer had tools and did not use them in time. Poor diligence does not legalize a fake list, but it does explain how a fake list made it through a $175 million approval process. About 350 JP Morgan people touched the deal. Axiom offered to do reverse phone lookups to test whether the numbers belonged to real people. The bank declined. The row count was treated as the test. If you take one investigative lesson from this case, take this. The breakthrough was not a hidden server, it was a sample. Somebody finally sent mail to the names on the page and counted what came back. If you cannot email a sample, you do not have users. You have a spreadsheet. On the deception spectrum, this case sits at the high end, not because Frank never existed, because the thing that made the price work was built for the buyer, not grown by the business. When Chase asked for proof, Javise didn't produce a customer file. She produced a model and a rented list and treated the row count as if it were a user count. That's not sloppy marketing, that's manufacturing evidence to close a deal. Each episode, I share a lie detector tip, a field tested lesson from real investigations to help you spot fraud before you become a victim. Today's tip concerns unverified scale. A large number is a representation, it is not evidence, until it can be independently confirmed. In financial crime, the damage often begins when a buyer, an investor, or a partner accepts a figure because it's convenient. Not because it's been tested. Three points matter. First is definition. The term being counted must be fixed in writing. If the word can mean one thing in a pitch and another in the records, the figure is unreliable. Second source. Determine whether the number was produced in the ordinary operation of the business or sampled for the transaction. Those are not the same thing. Third, access. You must be able to examine a sample without relying solely on the party that benefits from the number. If that access is denied, the representation has not been verified. Objections like timing, privacy, and prior review are inevitable. Some of those concerns are legitimate. They are also the most common barriers placed around a number that will not survive inspection. From a free federal form at a Wharton education to four million names that were never users in eighty five months in federal prison. Charlie Javiz did not simply overstate a number in a pitch. When the bank demanded a customer file, she had one manufactured. If this episode hit home, if you've ever been shown numbers that were already validated and not allowed to inquire further, I want to hear what happened. Send it to Kevin at Follow The Lives Podcast dot com. The most interesting notes may become future episodes. If you want practical tools to protect yourself from deception, download my free guide. Seven proven strategies to spot lives and uncover the truth in any interview. It's at Follow The Lives Podcast dot com. While you're there, join my substack at follow the lives. com for deeper dives and extra lie detector tips. If you're getting value from these episodes, subscribe and turn on notifications. On Apple Podcasts, our rating takes only seconds and it makes a difference. Thanks for listening. Until next time, follow the Lies.