The $85 Million Reconciliation Gap

You deposit $280,000 from the sale of your house into a fintech app. The app shows your balance. The app is not a bank, but it tells you the money is held at an FDIC-insured institution.

Six months later, the company between the app and the bank goes bankrupt. Your balance freezes. A court-appointed trustee tries to reconcile the records and finds that $85 million is missing across all accounts. Nobody can agree on where it went.

The app still shows your balance. The money behind it is gone.

That's Kayla Morris's story, and it's the central fact of the Synapse Financial Technologies collapse. Not a hack. Not a theft, at least not in the conventional sense. A sub-ledger that drifted from reality until the drift consumed $85 million in customer funds and left a bankruptcy court to sort out who owned what.

The architecture failure here is embedded in how banking-as-a-service works, and the fix isn't obvious.

The three-party ledger problem

Synapse sat in the middle of a three-layer stack that powered nearly 100 fintech apps.

At the top: the fintech brand. Yotta, Copper, Juno, and dozens of others. These companies built the app, ran the marketing, owned the customer relationship. They didn't hold any money.

In the middle: Synapse. It maintained the customer-level accounting, routed payments, handled compliance workflows, and connected the brands to the banks underneath. The sub-ledger lived here. All customer balances and the allocation of pooled funds to individual accounts ran through Synapse's systems.

At the bottom: the partner banks. Evolve Bank & Trust, AMG National Trust, American Bank, and Lineage Bank. They held the actual deposits in pooled accounts.

The pooled accounts are the critical piece. A partner bank doesn't open a separate account for each of Yotta's customers. It opens one account, an omnibus "For Benefit Of" (FBO) account, and Synapse's sub-ledger tracks which slice belongs to which customer. If 50,000 Yotta customers have a combined $40 million in deposits, Evolve holds $40 million in one account. The sub-ledger says Customer A owns $280,000 of it, Customer B owns $1,200, and so on down the list.

The bank sees one balance. The sub-ledger sees 50,000. If those two numbers don't match, someone's money isn't where the app says it is.

How the ledger drifted

In September 2023, Synapse flagged about $13 million in discrepancies between its records and the funds Evolve held. That same month, Mercury, Synapse's largest client representing roughly 35% to 60% of its revenue, moved more than $3 billion in deposits to work directly with Evolve, cutting Synapse out of the relationship. Evolve declared Synapse in breach of contract on September 25.

The relationship was already fracturing. Synapse accused Evolve of debiting fees from customer accounts that it wasn't entitled to. Evolve accused Synapse of maintaining unreliable ledgers. Both were pointing at the other entity's records as the source of the problem.

The constraint that made this catastrophic: neither party had an independent source of truth. The banks saw pooled account balances. Synapse saw customer-level allocations. No third party held a reconciled copy. When the two sides disagreed, there was no tiebreaker.

Synapse filed for Chapter 11 bankruptcy on April 22, 2024. A TabaPay acquisition fell through. The court removed management for "gross mismanagement" and appointed former FDIC Chair Jelena McWilliams as trustee on May 24. The last Synapse employees were fired the same day.

McWilliams inherited a company with no staff, no operating budget, and a sub-ledger that nobody who remained could interpret.

What the trustee found

The numbers told a simple, ugly story.

Fintech customers were owed approximately $265 million. The four partner banks held about $180 million associated with those accounts. The gap was $85 million.

McWilliams reported that Synapse had apparently comingled funds among several institutions, using multiple banks to serve the same fintech companies. She also noted that Synapse Brokerage, a subsidiary sharing employees and systems with the main company, may have affected fund flows, but the banks needed more information to determine how.

Evolve, meanwhile, said Synapse's ledgers showed large day-to-day balance swings with no corresponding fund movements. Phantom entries. Balances changing on Synapse's side without money actually moving at the bank.

The trustee described the situation plainly: the exact amounts and sources of the shortfall had not been clearly established. The estate had no funds to pay forensic accountants or investigate its own records. Evolve hired Ankura, a consulting firm, to attempt an independent reconciliation, but the process took months and produced figures that shifted as more data surfaced.

Think of it like a shared apartment where one roommate kept the only spreadsheet tracking who paid what share of rent. The roommate leaves. The landlord has bank statements showing total deposits. The other roommates have Venmo receipts. But the spreadsheet, the only record that connected each person's payments to the total, has errors nobody noticed until the person maintaining it was gone.

Except this apartment held $265 million.

The FBO architecture's single point of failure

The omnibus FBO account is not a flawed design. It's efficient, cost-effective, and operationally necessary when a bank serves thousands of customers through a single fintech partner. Opening individual accounts for every end user would be impractical at Synapse's scale of 18 million users.

But the design concentrates an enormous amount of trust in whoever maintains the sub-ledger. The bank can verify its own aggregate balance. It can confirm that $40 million sits in the FBO account. What it cannot independently verify is the allocation: which $280,000 belongs to Kayla Morris and which $1,200 belongs to someone else.

In a traditional bank, the institution that holds the money also maintains the customer records. The ledger and the cash are in the same building, under the same regulator.

When a middleware company operates the ledger on behalf of the bank, that link breaks. A company with its own incentives and its own engineering capacity holds the records that determine who owns what.

The FDIC published guidance in June 2024 clarifying what many customers had already learned: deposit insurance covers an insured bank failing. It does not cover a nonbank company's insolvency. Having FDIC-insured banks underneath the arrangement didn't protect Synapse's customers because the bank hadn't failed. The middleware had.

What reconciliation should look like

The gap between "how Synapse operated" and "how this should work" comes down to a daily engineering discipline.

A reconciliation system for FBO accounts needs to run on a cadence that catches discrepancies before they compound. At minimum, the sub-ledger total must match the bank's reported balance at the end of every business day. If Customer A deposits $500 and Customer B withdraws $300, the sub-ledger should show a net $200 increase, and the bank's FBO account balance should reflect the same $200 change.

When these numbers diverge, the system needs to halt outflows from the affected pool until the discrepancy is explained. An unexplained difference of $13 million, which Synapse identified in September 2023 and apparently continued operating through, should have triggered an immediate freeze on the affected accounts and a mandatory reconciliation with the partner bank.

Evolve said it "never" got Synapse to implement sufficiently disciplined reconciliation. Former employees described "crippling levels of tech failure" and a team of junior engineers with limited senior oversight. The company that maintained the only authoritative record of $265 million in customer allocations did not have the engineering capacity to keep that record accurate.

The FDIC proposed a rule in late 2024 that would require banks holding FBO accounts to maintain beneficial-owner records, essentially pushing the sub-ledger obligation down to the bank itself rather than trusting the middleware. The rule hasn't been finalized, but the direction is clear: the entity that holds the cash should also hold, or at least independently verify, the sub-ledger.

The legal aftermath

The reconciliation failure cascaded into a legal problem that is still playing out.

McWilliams gave the court two options: distribute funds proportionally to all customers, accepting that everyone takes a haircut, or fully pay some customers while making others wait for the reconciliation to finish. Neither option made anyone whole.

The bankruptcy case was eventually dismissed because the estate had no funds to support a Chapter 7 liquidation. The CFPB filed its own enforcement case in August 2025, securing a stipulated judgment in September 2025. The penalty was $1, a nominal amount that gave the agency access to its civil penalty fund to begin compensating harmed consumers.

In November 2024, Evolve began distributing funds and said it held $46 million ready for return plus $35.3 million in reserve. But the reconciliation remained incomplete. A federal grand jury investigation, dating to at least October 2024, examined potential accounting and wire fraud. Former controller Victor Medeiros received a subpoena. No charges have been filed.

Yotta sued Evolve in September 2024 for fraud and negligence. Mercury sued Synapse for $30 million in withheld revenue. Synapse countersued Mercury for $36 million. Lineage Bank's shareholders revolted over its Synapse relationship and ousted senior management.

Every party in the three-layer stack ended up in litigation with the others. The shared responsibility model meant nobody could resolve the problem without cooperation from the same companies they were suing.

What this forces you to build

If you're designing a BaaS platform or operating FBO accounts, the Synapse collapse reduces to a single engineering requirement: the sub-ledger cannot be a single point of failure.

That means automated reconciliation between the sub-ledger and the bank's aggregate balance every business day, with alerts that trigger when the variance exceeds a threshold.

It means the reconciliation data, the mapping between individual customers and their share of the pooled account, needs to exist in at least two independent systems. If the middleware goes down, the bank or a third-party escrow should be able to reconstruct every customer's position from its own copy.

And it means the sub-ledger operator's engineering capacity needs to be commensurate with the money it tracks. Synapse handled $76 billion in annual transaction volume with, according to former employees, an engineering team that lacked the seniority and tooling to maintain accurate records. $265 million in customer funds depended on a company that couldn't hire enough senior engineers to build a proper reconciliation pipeline.

The ledger is the product. The companies now applying for trust bank charters to bring custody and ledger-keeping under one roof are, in part, responding to this lesson. When the same institution holds the money and maintains the records, there's one entity to examine when they diverge.

Whether that accountability will satisfy regulators, courts, and the community bankers now suing the OCC over the charter path is a separate question. But the engineering lesson from Synapse is already settled.

If you can't reconcile your sub-ledger to the penny every night, you shouldn't be holding anyone's money.