A field guide · updated 2026
Embedded finance is why you can borrow money at checkout, get paid instantly for a delivery gig, or insure a flight without ever opening a banking app. This is a guide to how that actually works — the use cases, the companies making it possible, and what it means for the next decade of software.
You've almost certainly used it this month without noticing. If your platform already moves money for its users, the question is no longer whether embedded finance is relevant — only whether you've priced what you're currently giving away.
If you read nothing else
The rest of this page is the detail behind those five points, with sources.
01 — The definition
Embedded finance is the integration of a financial product — payments, lending, banking, insurance, or investing — directly into a non-financial company's product, at the exact moment it's needed.
The financial service isn't a redirect, a referral, or a link out to a bank's website. It lives inside the app the user already trusts, wearing that brand's interface, while a regulated financial institution operates quietly underneath.
This is a distribution shift as much as a technology one. For most of financial history, banks built the product and owned the relationship with the customer. Embedded finance splits those two jobs apart.
The practice is old — airlines and department stores issued their own credit cards decades ago. What's new is the term itself, which only entered common industry use around 2019–2020, once APIs made the "rent the plumbing" model fast and cheap enough for almost any company to try.
02 — Why now
Embedded finance isn't a new idea — store credit cards existed decades ago. What changed is that the underlying plumbing became rentable. APIs replaced years-long bank integration projects.
03 — Use cases
Nearly every embedded finance product falls into one of these categories.
Accepting or moving money without leaving the platform — checkout, marketplace payouts, in-app wallets.
Example: an ecommerce platform that lets merchants accept payments without a separate processor account.
Where the money lands: for payment and deposit products, McKinsey found revenue accrues mainly to the distributor that owns the end-customer relationship — the platform, not the bank. McKinsey, 2022
Financing offered at the point of decision, underwritten using data the platform already has.
Example: "Split into 4 payments" at checkout, or working-capital advances offered to sellers based on sales history.
Scale: writing in 2022, Bain forecast US BNPL transaction value of roughly $265B by 2026 against enabler and platform revenue of only about $4B — high volume, compressed margins. Treat as a directional model, not an outcome. Bain, 2022
Branded accounts, debit cards, and instant payouts issued through a partner bank, without the platform becoming a bank itself.
Example: gig-economy apps that pay workers into an in-app balance and issue a debit card against it.
Caution: this is the category where the 2024 Synapse collapse did its damage — pooled “for benefit of” accounts left over 100,000 end users locked out of their own balances. See risks →
Coverage offered inside the purchase flow it relates to, priced using the platform's own risk data.
Example: device or trip protection offered in the same flow as a purchase or booking, no separate application.
Why it works: the offer arrives at the moment the risk becomes salient to the buyer — the same policy sold weeks later, out of context, converts far less well.
Savings, round-ups, or investment products offered as a feature of a broader app rather than a standalone brokerage.
Example: a spending app that automatically invests spare change from everyday purchases.
Design note: the products that scale here remove the decision entirely. Anything requiring a deliberate contribution reverts to standard brokerage conversion rates.
04 — The ecosystem
It's tempting to picture embedded finance as "a brand plus a bank." In practice it's closer to a brand plus a bank plus a small crowd of non-bank fintech companies that never touch a dollar of the money but make the whole thing findable, sellable, and possible to integrate in the first place.
Every one of these parties is legally or technically involved in a given transaction.
The company the end user actually recognizes — a marketplace, an app, a piece of software they already use daily. Owns the user relationship and the interface.
The non-bank fintech that actually routes a transaction — authorizing, capturing, and passing it onward. Often the first "invisible" fintech company in the chain, and frequently the brand's main technical relationship.
Exposes banking functionality as an API — account creation, card issuing, ledgering, transaction handling — so the brand doesn't have to build a core banking system. Also a non-bank fintech, despite the name.
A chartered, regulated bank that actually holds the license to hold deposits and issue cards. Everything is legally happening at this bank, even though the user never sees its name. The one layer that has to be a real bank.
Visa, Mastercard, ACH, and real-time payment rails — the wires that actually move money between institutions once a transaction is authorized. The plumbing beneath the plumbing.
Every layer takes a sliver of the same transaction. On a typical card payment, the sponsor bank and card network split most of the interchange fee, the gateway and BaaS provider each take a per-transaction or per-account fee for the infrastructure, and the brand keeps whatever revenue share is left — often paired with a share of interest income on lending products. It's a thin margin on any single transaction, and a real business only at volume, which is exactly why embedded finance tends to favor platforms that already have a large, active user base rather than the smallest startups.
This is where most embedded finance startups actually live. None of these companies move funds themselves — they make the transaction stack above easier to build, choose, connect, or trust.
Connect a brand's app to a user's external bank accounts for verification, balance checks, or account-to-account payments — moving data, not money.
Sit above multiple BaaS providers or processors, letting a brand plug into several sponsor banks or payment routes through a single integration — mainly for redundancy and geographic coverage.
Identity verification, fraud scoring, and know-your-business checks that plug into the transaction stack at multiple points rather than owning any single layer.
Help brands evaluate and choose a BaaS provider, sponsor bank, or processor — demand-generation and trust infrastructure for the industry, closer to G2 than to a bank.
05 — How it works
A simplified version of what happens when a user taps "pay" inside an embedded finance product.
The brand's app sends an API call to its BaaS provider — for example, "charge this card" or "issue this payout."
KYC (Know Your Customer) and fraud checks run, often in milliseconds, using identity data collected when the user signed up.
The sponsor bank records the movement of funds on its regulated ledger — this is the legally authoritative record of the transaction.
Card networks or bank-to-bank rails (like ACH or real-time payments) actually move the money between institutions.
The result returns through the same chain, and the brand's app shows a simple confirmation — with none of the above visible.
Most of the engineering difficulty in embedded finance lives in the failure paths, not the happy path. A declined KYC check, a timed-out bank ledger call, or a network outage mid-transaction all need a defined, auditable outcome — money can't just be left in an ambiguous state. This is a large part of why "just call an API" still takes real engineering effort.
06 — The market
A non-exhaustive map of companies operating at each layer, by region. The stack looks similar everywhere; the companies filling it rarely do.
Companies operating across multiple regions, or whose infrastructure is used well beyond their home market.
The sponsor-bank model. Note that this segment went through significant consolidation and regulatory enforcement in 2024–2025 — see the risks section.
Structurally different: many providers hold their own EMI, PI, or full banking licence rather than renting a sponsor bank's charter.
Corporate history matters here more than in most regions. Solaris operates on a full German CRR banking licence, but discontinued major parts of the EMI business it had acquired as Contis in 2024, and previously operated under a BaFin restriction requiring approval before onboarding new partners. Railsr went through administration in 2023, was acquired, merged with Equals Money, and rebranded entirely as Equals in mid-2026 — the Railsr name no longer trades. Verify a provider's current licence category and ownership directly before shortlisting; the marketing site is not sufficient evidence.
Distribution here runs through super-apps and wallets more than vertical SaaS. Licensing differs substantially country by country, and several providers hold their own in-market licence rather than relying on a sponsor bank — a structurally safer arrangement than the US middleware model that failed in 2024.
Regional context: the ASEAN Regional Payment Connectivity initiative links domestic instant-payment systems — Indonesia's QRIS, Thailand's PromptPay, Singapore's PayNow, Malaysia's DuitNow, the Philippines' QR Ph and Vietnam's VietQR — which changes what cross-border embedded payments can look like here.
Built on public digital infrastructure (UPI, Aadhaar), with the RBI actively shaping what non-banks can and cannot do.
Brazil and Mexico are the most mature markets; infrastructure coverage thins considerably elsewhere in the region.
Mobile money is the dominant rail in much of the continent, and licensing expectations have tightened notably — Nigeria's central bank has pushed payment-only fintechs toward fuller licences.
Regulatory note: in early 2026 the Central Bank of Nigeria upgraded several fintechs including Moniepoint, OPay and PalmPay to national banking licences, while Flutterwave and Paystack have pursued microfinance bank licences. Licence category matters more than brand recognition when choosing a partner here.
Notice how few of these are actual banks. Most of the embedded finance industry — gateways, BaaS providers, aggregators, orchestration, compliance tooling — is non-bank fintech. The bank is one layer among many, not the centre of the picture. Company lists reflect publicly reported activity as of mid-2026 and change frequently; verify current licensing status directly before shortlisting.
07 — Where you operate
Almost everything written about embedded finance assumes the US sponsor-bank model. If you operate elsewhere, much of that advice doesn't transfer. This is the most common blind spot in the category.
A chartered bank holds the licence and the liability. There's no single federal regulator — the OCC, FDIC, Federal Reserve, and FinCEN each cover different pieces, and money transmission is licensed state by state, so cost and timeline vary widely by footprint.
Rather than renting a bank's charter, non-banks can hold their own Electronic Money Institution or Payment Institution licence under the PSD2/EMD2 framework. Authorisation in one member state passports across the EEA — a structurally different, more centralised route than the US.
Broadly similar EMI/PI licensing to the EU post-Brexit, but with the FCA's Consumer Duty adding an outcomes-based obligation that has to be demonstrated at product level, not just in policy documents.
Embedded finance in much of APAC grew through super-apps and wallets rather than vertical SaaS — a structurally different distribution model, with licensing regimes that differ substantially country by country.
"Do we need a bank partner, or can we hold our own licence?" has a different answer in Frankfurt than in Fresno. Any build-vs-partner decision has to start from the jurisdiction you actually operate in, and multi-jurisdiction operations multiply the compliance surface rather than adding to it.
08 — The honest part
Embedded finance is often pitched as pure upside. The 2024–2025 period proved otherwise, in public.
Synapse was a middleware provider connecting fintech apps to partner banks. When it filed for bankruptcy in April 2024, its ledgers turned out not to reconcile with what its partner banks actually held. More than 100,000 end users of downstream apps were locked out of their money. The court-appointed trustee reported roughly $265M owed to end users against about $180M held at partner banks. Reported shortfall figures vary by source and snapshot: the trustee cited about $85M, the CFPB later described a range of $60–90M, and other reporting has cited as much as $95M. What no source disputes is that the money could not be fully accounted for. CNBC / Banking Dive, 2024
The lesson leadership should take: the consumer-facing brands in that chain had made no promises they believed were false, and still absorbed the reputational damage. Ledger integrity two layers down your stack is your problem.
The Federal Reserve issued a cease-and-desist order against Evolve Bank & Trust, citing an ineffective risk-management framework for its fintech partnerships. The OCC placed Blue Ridge Bank under a consent order over Bank Secrecy Act and anti-money-laundering deficiencies. Enforcement actions have hit a long list of sponsor banks. Banking Dive, 2024
Blue Ridge had around 70 fintech partnerships at its peak. It exited banking-as-a-service entirely by the end of 2024, with its CEO stating plainly that volume had outrun the bank's capacity to run the programme properly. Every one of those partners needed a new home. American Banker, 2025
This is the single most repeated conclusion from the post-Synapse period. A brand offering embedded lending remains subject to consumer lending law; a brand offering deposits inherits obligations around how those funds are held and recorded, regardless of who operates the ledger.
Bain's 2022 BNPL model illustrates the pattern: roughly $265B in forecast transaction value producing about $4B in enabler and platform revenue, with explicitly compressed margins. Volume is not the same as profit. Bain, 2022
09 — What's next
Industry-specific software — for clinics, contractors, salons — increasingly adds payments, payroll, or lending tuned to that specific trade.
Embedded finance took off first in consumer apps. Business platforms — invoicing, procurement, freight — are next, with larger transaction sizes.
Platforms increasingly underwrite credit using their own behavioral and transaction data, not just traditional credit bureau data.
The counter-trend worth naming: some platforms that rushed into embedded lending or banking during the boom years are quietly scaling back, having learned that compliance overhead and balance-sheet risk aren't worth it below a certain scale.
10 — A gut check
Embedded finance isn't universally worth pursuing. Before scoping anything, it's worth answering four questions honestly. Editorial framework.
Interchange and revenue-share economics only work at scale. A few thousand transactions a month rarely covers the integration and compliance cost.
If people already pay, get paid, or hold a balance inside your product, adding a financial feature is a small step. If not, you're asking for a much bigger leap of trust.
Embedded finance means ongoing obligations — audits, reporting, dispute handling — not a one-time integration. That's a real, recurring operational cost.
Embedded finance isn't only an engineering project. It needs a named compliance owner, ongoing risk monitoring, and a support function equipped to handle people's money going wrong. If you can't name those people, you aren't ready to scope the build.
11 — The core decision
For most leadership teams this is the actual decision on the table. "Build" almost never means becoming a bank — it means how much of the stack you own versus rent. The "best when" and "main risk" columns are editorial judgment.
The honest default for most platforms is the first row. The interesting question isn't usually "should we own a bank" — it's "how exposed are we if our single partner exits the business."
12 — Measuring it
Embedded finance programmes fail quietly more often than they fail loudly. These are the numbers that tell you which way yours is going, well before the P&L does. Editorial framework — not an industry standard.
What share of eligible users actually take the financial product. Low attach usually means the offer is arriving at the wrong moment, not that the product is wrong.
The number that determines whether this is a business line or an expensive feature. Track it against the fully-loaded cost of the programme, not just the integration cost.
Include compliance headcount, support load, and dispute handling — not just per-account provider fees. This is the number most programmes underestimate at the business case stage.
Financial problems generate a different and much heavier class of support ticket than product problems. A rising rate here is an early warning that operational cost will outrun revenue.
Do users who adopt the financial product retain measurably better than those who don't? For many platforms this is the real return, and it's larger than the direct revenue.
13 — FAQ
Embedded finance is the integration of financial services, like payments, lending, or insurance, directly into non-financial products and platforms. Instead of visiting a bank, you get a loan at checkout, a debit card inside a delivery app, or insurance while booking a flight.
Fintech refers to technology-driven financial companies that sell financial products directly to consumers or businesses. Embedded finance is when a non-financial company embeds those same financial capabilities inside its own product, often powered by fintech infrastructure behind the scenes.
Banking as a Service is the infrastructure layer that lets non-bank companies offer banking products. BaaS providers connect a brand's app to a licensed, regulated bank through APIs, handling account creation, card issuing, ledgering, and compliance.
In most jurisdictions, only chartered banks can hold deposits, move money, or issue certain financial products. Non-bank brands partner with a sponsor bank, which holds the regulatory license, while the brand and BaaS provider handle the user experience and technology.
Common examples include buy-now-pay-later at ecommerce checkout, ride-share driver debit cards and instant payouts, insurance offered when booking travel, and invoicing platforms that let small businesses accept payments or access working capital directly inside the software they already use.
No. Open banking is about securely sharing bank account data with third parties, typically for account aggregation or payments initiation. Embedded finance is broader: it is about delivering entire financial products, like accounts, cards, loans, or insurance, inside a non-financial platform. Open banking is often one of the technical building blocks embedded finance relies on.
Ecommerce, ride-sharing and delivery, vertical software (SaaS for specific industries like healthcare or construction), travel, and gig-economy platforms are among the heaviest adopters, since they have large user bases with clear, recurring financial needs.
Key risks include regulatory complexity across jurisdictions, compliance liability that can extend across the brand, BaaS provider, and sponsor bank, dependency on a small number of sponsor banks, and the reputational risk a brand takes on when it starts to look and act like a financial institution.
Costs vary widely by product type, but most programs involve three cost layers: a setup or integration fee from the BaaS provider, ongoing per-account or per-transaction fees, and internal engineering and compliance time. Simple embedded payments can launch relatively cheaply; embedded lending or banking programs cost significantly more due to the compliance work involved.
It depends on the product and the sponsor bank, but compliance review is typically the longest part of any launch timeline, often taking longer than the technical integration itself. Simpler payment products move faster; anything involving lending, deposits, or credit underwriting requires more extensive review.
Most companies in the embedded finance ecosystem are non-bank fintech startups, not banks. Payment gateways, Banking-as-a-Service providers, account aggregators, orchestration platforms, and compliance tooling are all typically built by fintech companies. The chartered bank is only one layer in the stack; the rest of the ecosystem is software companies that never directly hold a banking license.
Yes, substantially. The US sponsor-bank model, where a chartered bank holds the licence on a platform's behalf, is not how the market works everywhere. In the EU and UK, non-banks can obtain their own Electronic Money Institution or Payment Institution authorisation, and an EU authorisation passports across the EEA. Much of Asia-Pacific developed through super-app distribution with country-by-country licensing. Advice written for one jurisdiction often does not transfer to another.
Synapse was a middleware provider connecting fintech apps to partner banks. Its 2024 bankruptcy revealed that its ledgers did not reconcile with what partner banks actually held, and more than 100,000 end users were locked out of their funds. The trustee reported roughly $265M owed against about $180M held, with shortfall estimates ranging from $60M to $95M depending on source and snapshot. It matters because it established, publicly, that consumer-facing brands absorb the reputational damage for failures deep in their stack, and it triggered a wave of regulatory enforcement against sponsor banks.
Sources
Market-sizing forecasts in this category vary widely by analyst and by what's being counted — transaction value and revenue are very different numbers. Figures here are attributed to their original source and dated, so you can judge them yourself.
A note on what these numbers are. The headline market-sizing figures on this page are analyst forecasts, most made in 2022 for a 2026 endpoint that has now arrived, and published before the 2024 sponsor-bank enforcement wave that reshaped the sector. No public actuals confirm whether they landed. They are useful for order-of-magnitude reasoning and nothing more precise than that. Where a number could not be sourced, it was left off this page rather than estimated. The frameworks in the metrics and build-vs-partner sections are editorial judgment, not industry standards, and are labelled as such.