Why rails matter
A "payment" is a customer-facing abstraction. A rail is the actual network the money rides on. Picking the right rail for each use case is one of the highest-leverage decisions a fintech makes — it sets cost, speed, reach, fraud profile, and reversibility characteristics all at once.
The major US rails at a glance
| Rail | Speed | Cost | Reversibility | Reach |
|---|---|---|---|---|
| ACH | T+1 to T+3 (batch) | Pennies | Yes (Nacha returns) | Every US bank |
| Same-day ACH | Same day, capped | Slightly higher | Yes | Every US bank |
| Wire (Fedwire) | Same day | $5–$30 | No (rare exceptions) | Every US bank |
| RTP | Seconds | Few cents | No | RTP-connected banks |
| FedNow | Seconds | Few cents | No | FedNow-connected banks |
| Card networks | Seconds (auth), T+1 settle | 1.5–3% (interchange + fees) | Yes (chargebacks) | Global |
| SWIFT | Hours-days | $25–$50+ | Partial | Global correspondent network |
Common pitfalls
- One-rail strategies. Every rail has outages and limits. Mature programs run a rail portfolio and route by use case.
- Confusing rail with experience. "Instant" is a UX claim; the rail underneath may be RTP, FedNow, or a card push.
- Ignoring corridors. International rails (SWIFT, correspondent, stablecoin) are corridor-specific. Don't generalize from one route to another.
Pick rails the way a logistics company picks shipping methods: by lane, not by default.