Everything in this course so far has assumed you are choosing how to move money for the sellers on your platform. The last decision is who builds the machine that does it. You can assemble settlement, the ledger, the payout engine, and the compliance plumbing yourself on raw acquiring and banking rails, or you can rent a platform that has already wired those pieces together. This lesson is about that choice, the three vendors most teams actually evaluate, and the cost that almost nobody prices correctly: leaving.
What "platform" really buys you
When you adopt a platform product, you are not buying a payment processor. You are buying the integrated version of the prior nine modules: split settlement, the merchant-of-record posture, seller onboarding and KYC, the multi-account ledger, the payout engine, and the multiparty compliance surface, all behind one integration.
The value is that you stop owning the seams between those pieces. The cost is that the platform now owns your data model. Your sellers exist as objects in their system, your balances live in their ledger, and your money movement obeys their state machine.
That trade is usually correct early. It stops being obviously correct once your volume is large enough that a few basis points and a few product constraints translate into real money and real roadmap drag.
The three you will evaluate
Stripe Connect
Connect is the default for platforms that want to be live this quarter. It offers Standard, Express, and Custom connected accounts that trade seller-facing control for onboarding speed and integration depth. Standard hands the seller a full Stripe relationship and dashboard; Express and Custom let you own more of the experience while Stripe still carries the underlying account.
The pricing shape matters for the build-vs-buy math. Standard accounts carry no Connect-specific platform fee on top of processing. Express and Custom add a per-active-account monthly charge and a per-payout fee on top of the base processing rate (Stripe lists roughly $2 per active account per month and about 0.25% + $0.25 per payout for those tiers). Those numbers move by region and over time, so confirm current rates against Stripe's own pricing page before you model anything.
Connect's strength is documentation, embedded onboarding components, and the fact that your engineers have probably used it before. Its constraint is that you are inside Stripe's opinions about how a platform should work.
Adyen for Platforms
Adyen sells to the enterprise end of the same problem. Its balance platform models seller funds as balance accounts, splits incoming funds between you and the seller at the point of sale, and runs payouts through that structure rather than a separate payout layer bolted on.
The differentiator is the embedded financial stack sitting next to payments: Capital (financing offered to sellers off their payments history), Accounts, and Issuing (platform-branded cards). Adyen reported its platforms business served 28 platforms with over €1 billion in annual payment volume at the end of 2024, up from 18 the year prior, so this is a real but still concentrated book, weighted toward larger customers.
Adyen tends to win when you want one acquirer across many markets, in-house card issuing, and a single regulated counterparty. It tends to lose on time-to-first-transaction against Stripe, because the integration is heavier and the onboarding is more enterprise sales than self-serve.
Mangopay
Mangopay is built around a programmable e-wallet and is the most "marketplace-native" of the three in posture. It is authorized as an Electronic Money Institution, by the CSSF in the EU and by the FCA in the UK (FCA EMI authorization came in November 2023), and it leans into European multiparty flows with customers like Vinted, Chrono24, and Rakuten.
The e-money wallet is the point. Funds sit in named wallets per seller, which makes hold-then-release flows, delayed payouts, and escrow-style timing first-class rather than something you simulate on top of a processor. For platforms whose model depends on holding money between pay-in and payout, that maps cleanly to the regulatory reality of holding client funds in Europe.
Mangopay's constraint is the mirror image of its strength: it is modular and EU/UK-centric, so a US-first platform or one that wants a single global acquirer will find the fit weaker.
A worked example
Take a services marketplace doing $40 million a year in gross seller payments across 5,000 active sellers, with weekly payouts.
On Connect Express, the platform-specific layer alone is roughly 5,000 active accounts times $2 a month, about $120,000 a year, plus per-payout fees of roughly $0.25 across roughly 260,000 payouts a year, another $65,000 or so, before any processing spread. That is around $185,000 a year in Connect-specific cost on top of interchange-plus, ignoring FX, refunds, and disputes.
The build alternative is not free, and this is where teams deceive themselves. To replicate that on raw rails you need the ledger from module 5, the payout engine from module 6, the onboarding and KYC from module 4, and the compliance surface from module 8, plus the headcount to operate them and the audit posture to satisfy a regulator. Two or three engineers and a compliance hire is comfortably north of $185,000 fully loaded. The platform fee is often cheaper than the team it replaces, right up until your volume makes the basis points dominate.
The switching cost nobody budgets
The real number is not the monthly fee. It is the cost of leaving once you are in.
The lock-in is the seller graph. Every connected account, balance account, or wallet is an object that lives in the vendor's system with their identifiers, their KYC verification state, and their payout history. Migrating means re-onboarding sellers somewhere else, which means re-collecting documents, re-running KYC, and re-establishing payout instructions, while money is in flight.
Three things make that hard in practice. KYC verification status rarely transfers between providers, so most sellers re-verify from scratch. Held balances and in-flight payouts have to be drained or reconciled across two ledgers during the cutover. And any stored card credentials on file with the old provider usually cannot be exported, so card-on-file sellers and buyers may need to re-authorize.
You price this by asking one question before you sign: if we had to move every seller to a new provider, how many would we lose, and how long would the dual-running period last? If the honest answer is "most of them" and "a quarter," the platform fee is not your real cost. Re-onboarding your entire seller base is.
Takeaway
Buy the platform when speed and a small team are worth more than basis points and control, which is true for most platforms most of the time. Pick the vendor by posture, not by headline rate: Connect for speed and self-serve breadth, Adyen for enterprise scale and embedded finance, Mangopay for European e-money wallet flows where holding funds is the model. And before you commit, write down the cost of leaving in sellers and quarters, not just in fees, because that is the number that will actually constrain you later.