In this post-Synapse analysis, payments infrastructure expert Vladyslav Kolodistyi maps how the Banking-as-a-Service architecture has been rebuilt since the 2024 collapse, and explains what every Embedded Finance team needs to verify before choosing a BaaS partner in 2026.
BaaS architecture before and after Synapse. The pre-2024 fragile stack with BaaS-owned ledger versus the 2026 resilient stack with bank-owned ledger. Architecture comparison by Vladyslav Kolodistyi.
The 2024 collapse of Synapse Financial Technologies was the single largest inflection point in the modern history of Banking-as-a-Service. Over 100,000 end customers lost access to roughly $265 million held across fintech platforms that depended on Synapse for ledger and reconciliation services. The collapse exposed an architectural weakness that almost every early-stage Banking-as-a-Service stack shared. Two years later, the Embedded Finance industry has rebuilt around a fundamentally different BaaS architecture, and the platforms that did not adapt are the ones now struggling for partner banks and regulatory approval.
Having worked in payments infrastructure throughout the post-Synapse rebuild, I have seen the architectural shift play out across dozens of Embedded Finance clients. The platforms that survived and the platforms that grew were the ones that re-architected their Banking-as-a-Service relationships around bank-owned authoritative ledgers, direct customer identification at the sponsor bank, and explicit FDIC pass-through coverage in writing. The platforms that kept the old fragile BaaS architecture have spent the past two years quietly re-papering contracts, switching partners, and rebuilding trust with regulators.
What Synapse's Collapse Taught the Embedded Finance Industry
The Synapse failure modes were not idiosyncratic. They were structural to the pre-2024 Banking-as-a-Service architecture. The BaaS middleware company held the authoritative ledger of end-customer balances. The sponsor bank held only aggregated For-Benefit-Of accounts and could not independently verify individual customer balances. When reconciliation between the BaaS ledger and the bank's view broke down, neither party could establish the truth quickly enough to protect customers. End customers were caught between a BaaS in financial distress and a bank that did not know them as individuals.
As Vladyslav Kolodistyi notes, the structural lesson was that the Banking-as-a-Service middleware should never own the authoritative customer ledger. The sponsor bank should own the ledger and see every end customer directly. The BaaS layer should be a software relay that orchestrates the consumer-facing experience without controlling the underlying record of who owes what to whom. The 2026 Embedded Finance stacks that work all share this architecture. The stacks that do not share it cannot get sponsor banks to onboard them in the current regulatory environment.
Vladyslav Kolodistyi explains that the post-Synapse Embedded Finance architecture also separates regulatory roles more clearly. The sponsor bank owns the compliance burden for the end customer. The BaaS provider owns the software burden for the fintech client. The Embedded Finance platform owns the customer experience burden for the end user. Each party has a clear remit. Each party can be held accountable. Each party knows what the other parties are doing without depending on opaque internal systems. That separation is what makes the 2026 Embedded Finance architecture resilient in ways the pre-2024 architecture was not.
"The structural lesson from Synapse was that Banking-as-a-Service middleware should never own the authoritative customer ledger. The sponsor bank should own it and see every end customer directly."
By Vladyslav Kolodistyi
The New Banking-as-a-Service Governance Stack After 2024
The regulatory response to Synapse reshaped how Banking-as-a-Service oversight works in practice. The OCC and FDIC tightened guidance for sponsor banks running Embedded Finance programmes. Sponsor banks now face stricter capital requirements when offering
BaaS services, more rigorous third-party risk management programmes, and direct supervisory attention on their BaaS books. The downstream impact is that BaaS providers face higher bars to even secure a sponsor bank, and Embedded Finance platforms in turn face higher bars to secure a BaaS partner.
Sponsor bank capital and concentration limits: banks now cap the share of total deposits that come through BaaS programmes, limiting how many fintechs any one bank can serve.
Independent reconciliation audits: third-party verification of ledger accuracy is now standard, replacing self-reporting by the BaaS provider.
Wind-down and resolution planning: every Banking-as-a-Service contract now includes documented procedures for returning customer funds if the BaaS provider fails.
Direct customer identification: end customers must be identified to the sponsor bank, not aggregated into FBO accounts where the bank cannot see individuals.
FDIC pass-through documentation: explicit written confirmation that each end customer's funds are covered up to the $250K insurance limit, with clear paths for proving coverage during a wind-down.
The eight non-negotiable criteria for choosing a Banking-as-a-Service partner in 2026, mapped to the specific Synapse failure modes each criterion addresses. Framework by Vladyslav Kolodistyi.
The eight criteria mapped in the framework above are what serious Embedded Finance teams now verify before signing any Banking-as-a-Service contract. According to Vladyslav Kolodistyi, every one of these criteria addresses a specific failure mode that contributed to the Synapse collapse. Skipping any criterion creates a verifiable governance gap that regulators will flag and that sponsor banks will increasingly refuse to onboard around. The criteria are not optional. They are the entry ticket for modern Embedded Finance partnerships, and Vladyslav Kolodistyi sees them being applied to every new sponsor bank conversation in 2026.
"Every one of the eight 2026 BaaS criteria addresses a specific Synapse failure mode. Skipping any criterion creates a governance gap that regulators flag and sponsor banks refuse to onboard around."
By Vladyslav Kolodistyi
Vladyslav Kolodistyi on Choosing a Resilient Banking-as-a-Service Partner
The practical work of choosing a Banking-as-a-Service partner in 2026 is harder than it was in 2022, and the harder process produces dramatically more resilient Embedded Finance outcomes. The Embedded Finance teams that I see succeeding follow a four-step BaaS partner evaluation: architectural review first, regulatory review second, operational review third, and commercial review last. The order matters. Teams that lead with commercial terms and reverse into architecture end up locked into BaaS contracts that cannot pass post-Synapse regulatory expectations.
- Architectural review verifies the BaaS partner uses the bank-owned ledger model. Walk through the data flow. Confirm the BaaS layer does not maintain the authoritative balance record. Confirm end customers are identified at the bank.
- Regulatory review confirms the sponsor bank is operating under the post-2024 BaaS oversight regime. Check the primary regulator, the recent supervisory letters, and the bank's history with previous BaaS programmes.
- Operational review tests the BaaS partner's actual reconciliation, audit, and dispute resolution processes. Run hypothetical wind-down scenarios. Verify the documented procedures exist and work.
- Commercial review negotiates the final BaaS terms once the architectural, regulatory, and operational reviews have passed. Skipping straight to commercial discussions creates the conditions Synapse exposed.
Embedded Finance platforms that follow this evaluation order ship resilient Banking-as-a-Service architectures and pass regulatory scrutiny when it comes. Platforms that skip the early reviews end up making the same architectural choices that Synapse made, and they will face the same kinds of risks Synapse faced once the next stress event hits the Embedded Finance industry. The next stress event is coming. The platforms with bank-owned ledgers, direct customer identification, and documented wind-down plans will navigate it. The platforms that did not rebuild after 2024 will not.
I write about Banking-as-a-Service architecture, Embedded Finance governance, and resilient payment infrastructure regularly. Find me on LinkedIn for the next analysis on BaaS partnership design.
Vladyslav Kolodistyi
Top comments (0)