Which payment rail for which B2B flow, in 2026

For most of the last two decades, the “which rail” question in US B2B payments barely existed — it was ACH or a paper check, and the check was losing. That is no longer true. There are now four serious ways to move money between businesses, each with a distinct cost, speed, limit, and reversibility profile, and the gap between them is wide enough that defaulting every payment onto one rail is a decision you are making by accident — usually the wrong one.

This is the framework we walk growth-stage teams through: stop asking “what rail do we use” and start asking “what does this specific flow need,” then route to the cheapest rail that meets it.

The four trade-offs that decide every flow

Every rail decision comes down to how a payment scores on four axes:

• Cost. Standard ACH is the cheapest bank-to-bank rail in the US at roughly $0.20 to $1.50 per transaction with no percentage fee, and is often bundled into business banking pricing at no per-item cost (Eco, 2026). Wires run $15 to $50; card and instant-to-card carry interchange typically in the 1 to 3 percent range, which compounds fast at scale (Routable, 2026).
• Speed. Standard ACH settles in one to three business days; same-day ACH can clear the same business day if submitted before a cutoff window; RTP and FedNow settle in seconds; wires settle same business day (BankXLSX, 2026).
• Per-payment limit. Same-day ACH is capped at $1 million per item today, rising to $10 million on September 17, 2027 (Beancount, 2026). RTP carries a $10 million ceiling, raised from $1 million in February 2025; FedNow defaults to $100,000, raisable to $500,000 by participating banks (Eco, 2026).
• Reversibility. ACH and same-day ACH can be returned, which is a feature when you need protection against error or fraud. Wires, RTP, and FedNow are final and irrevocable once sent — speed and finality come together, and so does the fraud exposure (BankXLSX, 2026).

The rails at a glance

Mind the cutoff windows

Same-day ACH is only “same day” if you make a window. The Federal Reserve and Nacha operate processing windows with submission cutoffs at roughly 10:30 a.m., 2:45 p.m., and 4:45 p.m. ET, and individual banks often close earlier (Relay, 2026). Miss the last window and a “same-day” payment becomes a next-day payment. Real-time rails have no such constraint — that 24/7/365 availability is precisely what you are paying for when weekend or after-hours settlement actually matters.

Which rail for which flow

Recurring, predictable payables → standard ACH

Vendor runs, payroll, subscription billing, and any scheduled disbursement where a one-to-three-day settlement is fine. This is ACH's home turf: cheapest cost, batch efficiency, and built-in reversibility if something is wrong. Do not pay a premium to accelerate a payment nobody is waiting on.

Time-sensitive but not instant → same-day ACH

An off-cycle supplier payment, a missed-invoice catch-up, a payroll correction — payments that need to land today but do not need to settle in seconds, and that benefit from keeping ACH economics and reversibility. Just build the cutoff windows into your operations so “same day” is real.

Real-time, high-stakes, funds-on-confirmation → RTP or FedNow

Where releasing goods, services, or the next step depends on confirmed, final funds: instant supplier settlement, insurance and claims payouts, real estate and closing flows, earned-wage access, and marketplace payouts that are a competitive feature (Payments Dive, 2026). Choose RTP for larger payments up to $10 million; use FedNow where reach or your bank's participation favors it, within its lower default limit. The controls must be right first — these do not reverse.

High-value, one-off, final → wire

Large, infrequent, must-settle-today payments above the instant-rail limits — an acquisition payment, a large capital transfer. You pay for it, but the finality and the very high ceiling are the product.

Supplier payables where data and rebate win → virtual card

When the interchange cost is outweighed by rebate economics, controlled single-use credentials, and rich reconciliation data, a virtual card can turn accounts payable from a cost center into a revenue line — provided the supplier accepts cards. That acceptance-and-economics design is its own decision, which we break down in our virtual card AP economics teardown.

Why this is a routing problem, not a rail choice

The reason companies leak money here is that they treat rail selection as a one-time setup rather than a routing rule applied to every payment. Once you accept that different flows want different rails, the real work is the logic that decides — automatically — which payment goes down which rail based on its urgency, value, reversibility need, and the receiver's capabilities. That is multi-rail routing, and it is where both the savings and the control live. It also connects directly to the economics on the card side: if a meaningful share of your volume runs through cards, it is worth stress-testing your processor and network path in the same exercise, because the markup there compounds with every transaction.

FAQ