SWIFT gpi and the Myth of Real-Time Trade Payments

SWIFT launched gpi in 2017 with a clear promise: cross-border payments would be fast, transparent, and trackable. By 2025, SWIFT reported that over 50% of gpi payments were credited within 30 minutes, and nearly 90% within 24 hours. Those numbers are real.

They are also irrelevant to anyone working in trade finance.

A treasury manager at a European commodity trading house told me last month that their average trade finance payment, from document approval to beneficiary credit, takes four to six business days. On complex corridors (West Africa, Central Asia, parts of Southeast Asia), seven to ten days is common. "I see the gpi dashboard. I see the tracker. I know exactly which correspondent bank is sitting on my payment at 3am. Knowing where it is stuck does not make it move faster."

What actually happens to a trade finance payment

To understand why gpi's headline statistics do not apply, walk through a real payment chain.

A confirming bank in Frankfurt needs to pay an exporter in Ho Chi Minh City against an LC presentation for a $4.2 million shipment of rubber.

Internal approval (1-3 days). The payment instruction cannot be generated until the trade operations team confirms documentary compliance, any discrepancy waivers are processed, and the payment is authorized against the credit line. At a mid-sized European bank, according to a former trade operations head, this step alone averages two business days. At larger banks with layered approval workflows, three days is not unusual.

Compliance screening (hours to 2 days). The payment instruction hits the sanctions filter, AML checks, and, for this corridor and commodity type, enhanced due diligence. A $4.2 million payment for rubber from Vietnam to a German counterparty crosses multiple screening thresholds. Manual review by a compliance analyst is triggered. That analyst has a queue.

SWIFT gpi does not track any of this. The timer starts when the payment instruction enters the SWIFT network. Everything that happens inside the bank before that moment is invisible to the tracker.

The correspondent chain (1-2 days). Frankfurt to Ho Chi Minh City routes through at least two intermediaries: typically New York for USD clearing and Singapore or Hong Kong as a regional hub. Each correspondent bank applies its own compliance screening and operates on its own cut-off schedule.

Cut-off times are the invisible friction. A payment operations specialist at a US correspondent bank described the dynamic: "If a payment instruction lands at 4:45pm and our cut-off is 4:30pm, it processes next business day. Nobody did anything wrong. The payment just lost 12 hours because of 15 minutes."

Beneficiary credit (0-2 days). The receiving bank in Vietnam credits the exporter's account. For large inbound transfers, the bank may apply a hold pending its own internal review.

Total: four to eight business days for a routine corridor. The gpi tracker covered the one to two days in the middle, the time the payment actually spent on the SWIFT network. The rest of the delay happened before and after the tracker could see it.

Why aggregated statistics mislead

SWIFT's published gpi metrics are averages across all payment types. Retail remittances and standard corporate payments, which account for the overwhelming majority of gpi volume, are high-frequency, low-value, and route through well-established corridors with pre-screened counterparties. They settle fast, and they dominate the average.

Trade finance payments have the opposite profile: low frequency, high value, non-standard routing, and compliance-intensive. They are a small fraction of gpi volume, which means their slower processing times barely register in the aggregate statistics.

SWIFT does not publish a separate gpi benchmark for trade finance payments. Several bankers I spoke with said they have asked for one. "We raised it at the gpi member group meeting," said a payments head at a bank in the Middle East. "The response was that they are 'exploring segmented reporting.' That was over a year ago."

What gpi actually fixed

The visibility improvement is genuine and should not be dismissed. Before gpi, a payment sent from Frankfurt to Ho Chi Minh City entered a black box. Queries to correspondent banks were sent by SWIFT MT199 message and could take days to receive a response. Exporters waiting for payment had no information. Banks had no diagnostic capability.

The gpi tracker changed that. Banks can now see which correspondent is holding a payment and for how long. The operations team at the Frankfurt bank can call the New York correspondent and ask why the payment has been sitting for 18 hours. That conversation happens in real time instead of over days of SWIFT messages.

Fee transparency was the other meaningful improvement. Before gpi, each correspondent in the chain deducted its processing fee from the payment amount. The beneficiary received less than expected with no visibility into the deductions. Under gpi rules, fees are disclosed upfront. For trade finance, where the exporter expects to receive the exact LC amount and any shortfall triggers a dispute, this matters.

What would actually help

Separate SLAs for trade finance. Measuring trade finance payment performance against a benchmark dominated by retail remittances is meaningless. A dedicated three-business-day SLA for trade finance payments, measured from document approval to beneficiary credit rather than from SWIFT network entry to exit, would give the industry an honest metric. It would also force banks to address the internal bottlenecks that gpi currently cannot see.

Pre-positioned liquidity on key corridors. Several banks are experimenting with pre-funded nostro accounts at correspondent banks in high-volume trade corridors. When a payment is authorized, the funds are already in place. This eliminates correspondent processing time but requires locking up capital, a trade-off that only pencils out on corridors with predictable, high-frequency flows.

Parameterized compliance for repeat relationships. When the same two banks transact 200 times a year on the same corridor for similar commodity types, the 201st payment should not require the same manual review as the first. Building compliance models that recognize established patterns and apply proportionate screening would address the single largest source of delay. Some banks are piloting this internally. None have deployed it at scale.

gpi solved the visibility problem. It did not solve the speed problem, because the speed problem lives in compliance queues, internal approval workflows, and cut-off time mismatches that no messaging platform can reach. The pipes are faster. Everything around the pipes is not.

-- Tamara