Why Stripe Built Connect: A Product Story
A story about why platforms need primitives, not one merchant flow, and how to design for it.

We were selling to businesses. Then we tried to sell to a business that sells for others. Our neat merchant flow fell apart in a week.
What broke when we sold to sellers
The platform looked like any other prospect. Big traffic. Clear use case. We turned on the standard merchant flow. One account. One bank. One owner. Simple.
Then their team asked three things. First, how do we open accounts for each of our sellers? Second, how do we split fees across us and them? Third, who holds the bag when a dispute lands?
We had no good answer. Our onboarding asked for the platform’s details, not their sellers’. Our payout settings knew one bank, not hundreds. Our risk rules assumed one business, not many small ones under a bigger roof.
I went back to the docs of tools that had solved this class of problem. Stripe’s choice with Connect stood out: do not ship one merchant UX. Expose primitives.
What Stripe did differently
Connect put building blocks in front of platforms. Accounts for each seller. API-first onboarding. Transfers and fee routing. Payouts and reporting scoped per account. Clear lines on who does what.
- Separate duties: platform vs sub-merchant. Make it explicit in product and docs.
- Build for API-first flows, not one web form.
Why platforms are not big merchants
| Assumption | Ordinary merchant | Platform or marketplace |
|---|---|---|
| User model | One business, one owner | Many businesses under one platform |
| Onboarding | Single KYC/KYB package | Tiered KYC/KYB across many sellers |
| Money movement | Settle to one bank | Collect, split, hold, route to many banks |
| Risk and disputes | One party liable | Shared or assigned liability across parties |
| Reporting | One ledger, one dashboard | Per-seller ledgers plus platform view |
| Support | Merchant support | Two-layer support: platform and seller |
How that choice changes product work
Onboarding becomes modular. You need a way to collect identity and business data per seller, with levels and triggers. KYC is no longer a fixed form. It is a state machine.
Liability needs clear rules in product. Who can refund. Who can dispute. Who holds reserves. The UI and APIs must reflect those lines.
Surface area explodes. Accounts, balances, transfers, payouts, capabilities, webhooks. You move from one dashboard to many views with scoped rights.
The payoff is real. Platforms adopt faster when they can compose their own flows. New revenue lines appear: onboarding services, payout features, reporting packages. The cost is real too: more regulatory work, more partner support, more docs.
Worked example: map the platform journey
Pick one platform customer. Start at their success moment. Work backward. A Business model canvas helps you frame value and partners, but here is the concrete build plan.
Write the end state in one line. Example: “Seller receives payout T+3 with correct fees and a clear statement.”
Identity, onboarding, fee routing, payouts, disputes, reporting. These are your product objects and APIs.
For each primitive, state who does what: platform vs sub-merchant vs provider. Reflect this in roles and scopes.
Write the webhook and state transitions. When is identity collected. When are capabilities granted. When do payouts unlock.
Set who absorbs disputes, who can refund, and reserve rules. Put this in product flags, not only in contracts.
For each primitive: owner, SLA, fallback. Keep it to one page. Share with sales, risk, support.
Mistakes I made so you do not
- Treating platforms like big merchants. It hides multi-party needs and creates rework.
- Hiding controls in docs. Put ownership and scopes in product, not footnotes.
- Forcing one onboarding flow. Offer levels and triggers tied to risk and volume.
- Under-resourcing compliance and support. This model needs both from day one.
What it taught me
Connect showed a path: ship primitives, define lines, and let platforms build. The point: when your customers serve others, do not sell a single UX. Sell building blocks and clear responsibilities.
You will move faster with the right customers. You will add new revenue. You will also take on more regulatory and product support work. Plan for that trade.


