Payment Processing Platform
Route, authorize and reconcile payments with idempotency and ledgering.
AdvancedCommerceFintech
Payment requests are idempotency-keyed and routed to the right processor/rail, with every state transition recorded in an append-only ledger so money movement is fully auditable and reconcilable against processor statements.
When to use it
- Payments must never be double-charged even under client retries or network failures
- Every movement of money needs a durable, auditable trail for reconciliation and disputes
Trade-offs
- Idempotency keys, ledger entries and reconciliation jobs add real implementation weight beyond a naive charge call
- Multi-processor routing needs a fallback strategy for when a preferred rail is degraded
Components used
Web AppAPI GatewayManaged App ServiceCacheThird-party ClientRelational DatabaseBatch / Scheduled Job
How it works
- Every payment attempt carries a client-supplied idempotency key, so retries return the original result instead of charging again.
- Authorization and capture are separate steps: authorization reserves funds, capture actually moves them, which is why a refund and a void are different operations.
- A double-entry ledger records every movement as balanced debits and credits, and reconciliation compares that ledger against the processor's settlement reports.
Used in the wild
- Marketplaces splitting funds between platform and sellers.
- Subscription billing with retries and dunning.
- Any system where the finance team needs to explain a discrepancy to the penny.
Good to know
- Never store money as a floating-point number. Use minor units as integers or an exact decimal type — 0.1 + 0.2 is famously not 0.3 in binary floating point, and that error compounds across millions of rows.
- The ledger should be append-only. Correcting a mistake means writing a reversing entry, not updating the original, because auditors need to see that the mistake happened.
Related system design
E-commerce Checkout
Cart, inventory reservation, payment and order fulfilment as a saga.
Ride Matching (Uber-style)
Match riders to nearby drivers using live location and a geo index.
Food Delivery Platform
Coordinate order, restaurant prep, and courier assignment end to end.
Event Ticket Booking
Reserve seats atomically to prevent double-booking under high concurrency.