I have built marketplace payments both ways in production: on Stripe Connect, and fully custom — gateway pay-ins, an internal double-entry ledger, and a disbursement rail paying sellers worldwide. Both are right, for different businesses at different stages. This guide is the decision framework I use, with the costs both paths hide.
What Connect actually sells you
Stripe Connect is not a payments API; it is an outsourced money services business. The fee line buys:
- Seller onboarding and KYC — identity verification, sanctions screening, and the compliance treadmill of keeping thousands of sellers verified across jurisdictions.
- Payout infrastructure— bank rails in dozens of countries, payout scheduling, failure handling, and the support burden when a seller's bank bounces a transfer.
- A compliance posture — money flows structured so the platform generally stays out of money-transmitter territory, with tax-form generation (1099-K et al.) handled per threshold rules that change yearly.
- Charge mechanics — destination charges, separate charges and transfers, application fees, and refund/dispute flows that correctly unwind split payments.
Founders comparing "Connect fees vs processor fees" are comparing a payments team's annual budget against a rate card. The honest comparison is Connect vs processor fees plus the ledger, the payout operation, the compliance counsel, and the two engineers who own it all.
When custom earns its complexity
- Volume with concentrated corridors. High GMV flowing through a few markets makes the per-transaction spread worth capturing, and few corridors keeps payout complexity bounded.
- Payment margin is the business. If your take rate lives inside the payment flow, owning the stack is owning your margin.
- Flows Connect does not model well — long escrow holds, milestone releases, split-and-recombine payments, wallet balances that sellers spend inside the platform.
- You already run the operational muscle — a team comfortable with reconciliation, disputes, and compliance review. The stack is buildable in months; the operation is forever.
The custom shape that works — the one I built for a creator marketplace processing thousands of transactions monthly — is three deliberately separate parts: an NMI-style gateway for pay-ins, an internal double-entry ledger as the single source of truth for every balance, and a MassPay-style disbursement provider for payouts through an FBO structure. Each part is replaceable because the ledger, not any provider, holds the truth.
The decision, compressed
- Pre-product-market-fit: Connect, almost without exception.
- Growing, standard flows, global sellers: Connect, renegotiate pricing at volume.
- High volume, few corridors, margin-critical, escrow-heavy: custom starts winning.
- Either way: keep your own ledger from day one. It is the option that keeps every other option open.
Deciding this for real money? I have shipped both architectures and reviewed many more. The clarity call pressure-tests your choice in 90 minutes; the platform engagement builds it.
Frequently asked questions
Is Stripe Connect worth the fees for a marketplace?
For most marketplaces, yes, for longer than founders expect. Connect's fees buy seller onboarding and KYC, payout infrastructure across dozens of countries, compliance posture, and fraud tooling — capabilities that cost far more to build and operate than their fee line suggests. The custom path starts winning when volume is high, sellers are concentrated in few markets, and payment margin is a core part of the business model.
What does a custom marketplace payment stack look like?
Three separately-chosen parts: a processor for pay-ins (an NMI-style gateway or a direct acquirer), your own double-entry ledger as the source of truth for balances and escrow, and a payout rail (a MassPay-style disbursement provider) usually structured around an FBO account. You gain margin and control; you own reconciliation, compliance exposure, seller verification, and every edge case Connect would have eaten.
Can I start on Stripe Connect and go custom later?
Yes, and it is usually the right sequencing: Connect to find product-market fit, custom when the unit economics justify a team owning payments. Design for it from day one by keeping your own ledger of marketplace balances even while Connect handles the money — the migration then swaps rails underneath a ledger you already trust, instead of reconstructing history.
Written by Andrej Dragojevic, Stripe Certified Professional Billing Architect.