Why Cross-Border Payments Are Still Expensive: The Last-Mile Problem

Send $200 from a bank account in London to a bank account in Lagos and, on average, about $12 of it disappears before it arrives. That is roughly 6% — double the 3% target that G20 governments set for global remittance costs back in 2021, and nowhere close to the sub-1% cost of moving the same $200 domestically through a real-time payment rail. The money isn't slower because the internet is slow. It's expensive because of what happens in the last few hundred meters of the journey, not the thousands of miles before it.

That gap between "how cheap this could be" and "how expensive this still is" is one of the most persistent puzzles in financial infrastructure. The core payment rails — SWIFT messaging, correspondent banking networks, card schemes — have had decades to mature. Yet the cost curve for cross-border payments has barely moved. Understanding why requires looking past the wire transfer itself and into the fragmented, locally-regulated, often manual process of actually getting money into someone's hands on the other end.

What Actually Happens When Money Crosses a Border

A cross-border payment is rarely a single transaction. It's a chain of intermediary steps, each one adding cost, delay, and risk.

When a bank in Country A doesn't have a direct relationship with a bank in Country B, it routes the payment through one or more correspondent banks that do have relationships on both sides. SWIFT — the messaging network — tells everyone in the chain what should happen, but SWIFT itself doesn't move money. Each correspondent bank in the chain:

• Takes a processing fee for handling the transaction
• Applies its own foreign exchange margin if a currency conversion happens on its books
• Runs its own sanctions and anti-money-laundering (AML) screening
• Settles with the next bank in the chain, often with a lag of hours to days

A payment that touches three or four correspondent banks accumulates three or four sets of fees and three or four rounds of compliance checks, none of which are visible to the sender or the recipient. This is sometimes called the "correspondent banking chain problem," and it's the reason cross-border payments have historically taken 2-5 business days even for corridors between two wealthy, well-banked countries.

The First Mile vs. The Last Mile

It helps to split the journey into two halves:

1. The first mile — moving money out of the sender's home banking system and converting it into an internationally transactable form. This part is increasingly fast and cheap. Modern payment processors, fintech aggregators, and improved SWIFT messaging (via the ISO 20022 standard) have compressed this stage significantly.
2. The last mile — getting the money out of the international system and into a recipient's local bank account, mobile wallet, or cash pickup point, in local currency, compliant with local regulation.

The first mile has gotten dramatically better over the last decade. The last mile has not, and it's where most of the cost and delay actually live.

Why the Last Mile Is So Hard

The last mile is expensive for structural reasons that don't disappear just because a company builds a slicker app.

Local payout infrastructure is fragmented and non-standardized. Every country has its own domestic payment rails, its own banking regulations, its own licensing requirements for who can distribute funds, and often its own preferred payout method — a bank transfer in one country, a mobile money wallet in another, physical cash pickup in a third. A payments company operating in 50 countries effectively has to build or license 50 different last-mile integrations, each with its own compliance regime.

FX conversion happens at the worst possible point in the chain. Currency conversion is often executed by whichever intermediary is last in the chain and closest to the local payout — precisely the point with the least competitive pressure and the least transparency to the end user. This is why the advertised "fee" on a remittance app is frequently a fraction of the real cost; the rest is baked into an FX spread that's never itemized.

Compliance is duplicated at every hop, and correctly so. Each jurisdiction in the chain has its own AML, counter-terrorism-financing, and sanctions-screening obligations. A bank in the middle of a chain can't simply trust that the sending bank already checked — it's legally required to check again. This isn't a solvable inefficiency; it's the intended function of a compliance regime built around jurisdictional sovereignty.

Local licensing gates who can even touch the money. In most countries, distributing funds to end recipients requires a local money transmission license, a banking partnership, or both. Global payment companies either acquire these licenses market by market (slow, capital-intensive) or partner with local players who take a cut for providing last-mile access.

Liquidity pre-funding ties up capital. Many payment providers avoid the delay of moving money hop-by-hop by pre-funding local accounts in destination markets — holding a float of local currency in Lagos, Manila, or Nairobi so that a payout can be instant on the receiving end. That float is capital sitting idle, and the cost of holding it gets passed back into the fee.

Why This Matters Right Now

The G20 set a target in 2021: get the global average cost of sending a $200 remittance down to 3% by 2030. The most recent data puts the actual global average closer to 6% — meaning a $200 transfer still loses around $12 to fees and spreads, roughly double what the target allows. That gap isn't a rounding error. Remittances are one of the largest flows of capital into developing economies, frequently exceeding foreign direct investment and official development aid combined for many recipient countries. A persistent 3-percentage-point gap between the target and reality represents billions of dollars a year that never reach the households the money was sent to support.

It also matters because the "first mile" half of this problem is genuinely close to solved. Sending money has never been easier from the sender's side — a phone, an app, a few taps. That makes the last-mile bottleneck more visible, not less, because the contrast between "friction-free to send" and "still expensive to receive" is now stark enough that customers notice it directly, usually in the form of a disappointing exchange rate on the receiving end.

Who Is Trying to Fix It, and How

A handful of structural approaches have emerged to attack the last-mile problem directly, rather than trying to make the first-mile experience marginally nicer.

Direct real-time payment system linkages

Instead of routing through correspondent banks, some countries are connecting their domestic real-time payment systems directly to each other. India's UPI linking with Singapore's PayNow is the most cited example of this model: a payment initiated in one country's real-time rail settles directly into the other country's real-time rail, skipping the multi-hop correspondent chain entirely. This collapses both the cost and the settlement time, but it only works between the specific countries that have built the bilateral (or multilateral) linkage — it doesn't generalize to an arbitrary corridor.

Pre-funded local payout networks

Fintech remittance and payout companies (rather than relying on ad hoc correspondent routing per transaction) maintain pre-funded accounts and local payment licenses in dozens of destination markets. The sender's money never actually crosses the border in real time; instead, the company nets transactions across its own book and releases already-present local currency to the recipient. This is faster and often cheaper than correspondent banking, but it requires enormous working capital and a license in every market served — which is why this model tends to consolidate around a small number of well-capitalized players.

Stablecoins and blockchain settlement rails

Stablecoin-based transfers attempt to solve the middle of the chain — moving value across borders without correspondent banks — by settling on a blockchain instead. This can meaningfully cut the first-mile-to-last-mile transit cost and time. It does not, however, solve the last mile itself: a stablecoin still has to be converted into local currency and paid out through a locally licensed on/off-ramp, which reintroduces most of the same licensing, compliance, and FX-spread dynamics described above. The last mile remains a local-currency, local-regulation problem regardless of what settles the middle leg.

Standardized messaging (ISO 20022)

The global shift from SWIFT's older MT message format to the richer ISO 20022 standard improves the quality and structure of payment data traveling with a transaction — reducing manual investigation of failed or flagged payments, which is itself a meaningful, if less visible, cost driver. Better data doesn't eliminate the last-mile licensing and FX problems, but it reduces the compliance friction layered on top of them.

Payment aggregators and orchestration platforms

Rather than building last-mile infrastructure themselves, many businesses now use payment orchestration platforms that route each transaction through whichever combination of local partners, licensed payout providers, or real-time rail is cheapest and fastest for that specific corridor at that specific moment. This shifts the fragmentation problem from "something a business has to solve" to "something an infrastructure vendor abstracts away" — at the cost of a margin taken by that vendor.

Practical Implications for Businesses and Builders

For a company that pays international contractors, receives payments from overseas customers, or runs payroll across borders, the last-mile problem shows up as a very concrete cost and reliability issue.

• The advertised fee is rarely the real cost. Compare total cost of a transfer — fee plus FX spread against the mid-market rate — not just the line-item fee, which is often deliberately minimized while the spread absorbs the rest.
• Corridor matters more than provider. The same payment provider can be excellent on a US-to-Mexico corridor (dense competition, good local infrastructure) and mediocre on a corridor into a smaller or less-banked market. Evaluate providers per corridor, not as a blanket choice.
• Speed and cost trade off differently by corridor. A corridor with a real-time payment linkage can be both fast and cheap; a corridor still dependent on multi-hop correspondent banking will usually be one or the other, rarely both.
• Pre-funded/local-payout models tend to beat correspondent banking on recurring payment flows (payroll, recurring vendor payments) because the provider can net transactions and avoid re-routing every single payment through the full chain.
• For one-off or unpredictable payments, correspondent banking or card-network rails may still be more practical, even at higher cost, because they don't require setting up a dedicated payout relationship in advance.
• Regulatory and licensing exposure is real if you build payment infrastructure yourself. Attempting to operate last-mile payout capability directly (rather than through a licensed partner) means acquiring money transmission licenses market by market — a slow and capital-intensive undertaking that most companies are better off avoiding by using an existing licensed provider.

A Simple Framework for Evaluating a Cross-Border Payment Route

1. Identify the corridor (sending country to receiving country) — cost and speed vary enormously by corridor, not just by provider.
2. Check whether a direct real-time payment linkage exists between the two countries; if so, it is very likely the cheapest and fastest option available.
3. If no direct linkage exists, compare the total cost (fee + FX spread against mid-market rate) of at least two providers with strong local payout presence in the destination market.
4. For recurring flows, prioritize providers with pre-funded local liquidity in the destination market over those relying on correspondent banking per transaction.
5. For compliance-sensitive or high-value transfers, weigh settlement time against the depth of AML/KYC screening a provider applies — faster is not always better if it comes with weaker screening in a higher-risk corridor.

Limitations and Open Questions

The last-mile problem is not purely a technology problem, which is why "just add blockchain" or "just build a better app" solutions consistently underdeliver against expectations.

• Licensing is a genuine barrier, not an inefficiency to be optimized away. Money transmission licensing exists to satisfy legitimate regulatory objectives around financial crime and consumer protection. Any solution that bypasses it isn't cutting cost — it's cutting compliance, which creates a different kind of risk.
• Real-time payment system linkages don't scale linearly. Each new bilateral or multilateral linkage (like UPI–PayNow) requires its own negotiation, technical integration, and regulatory alignment between the two countries involved. There is no shortcut to a fully connected global mesh of real-time rails; it has to be built corridor by corridor.
• Consolidation risk. Because pre-funded local payout networks require significant capital and licensing investment, the market has tended to consolidate around a small number of large players per region. That can reduce the competitive pressure that would otherwise push costs down further.
• FX spread transparency remains inconsistent. Even where regulation requires fee disclosure, the FX spread — often the larger cost component — is not always disclosed with the same rigor, making genuine cost comparison across providers harder than it should be for both consumers and businesses.
• Smaller and less-banked corridors lag furthest behind. Cost improvements have concentrated in high-volume corridors between well-banked economies. Corridors into markets with thinner banking infrastructure — often precisely the markets most dependent on remittance inflows — see the smallest cost improvements and the highest fees.

What to Watch Next

The trajectory of cross-border payment costs over the next several years will likely depend on a few specific developments:

• Expansion of real-time payment system linkages beyond the handful of pairs that currently exist, particularly involving major remittance-receiving markets in Africa, South Asia, and Latin America.
• Regulatory clarity on stablecoin settlement for cross-border value transfer, which would determine whether blockchain-based middle-mile settlement becomes a mainstream complement to correspondent banking or remains a niche alternative.
• Progress against the G20's 3% target, tracked publicly and used as the reference point for whether last-mile reforms are actually closing the gap or merely shifting where the cost is hidden.
• Continued rollout of ISO 20022 across major payment corridors, and whether richer payment data measurably reduces compliance-driven delays and failed-payment investigation costs.
• Consolidation or fragmentation among last-mile payout providers, which will shape whether competitive pressure keeps pushing fees down or whether a handful of dominant players in each region can hold pricing steady.

FAQ

Why are cross-border payments still so much more expensive than domestic ones?

Domestic payments typically move within a single, standardized national payment rail with one regulatory regime. Cross-border payments cross multiple banking systems, currencies, and regulatory jurisdictions, each of which adds its own fees, FX spread, and compliance checks — most of that cost concentrated in the "last mile" of local payout rather than the international transit itself.

What is the "last-mile problem" in payments?

It refers to the final step of a cross-border payment — converting funds into local currency and delivering them into a recipient's local bank account, mobile wallet, or cash pickup point. This step depends on local licensing, local infrastructure, and local compliance rules that don't standardize across countries, making it the most expensive and least efficient part of the journey.

Does using cryptocurrency or stablecoins make cross-border payments cheaper?

It can reduce the cost of the middle leg — moving value across borders without a correspondent banking chain — but it doesn't eliminate the last-mile problem. The stablecoin still needs to be converted to local currency and paid out through a locally licensed provider, which reintroduces most of the same FX and compliance costs.

What is the G20's 3% remittance cost target?

In 2021, the G20 set a global target to bring the average cost of sending a $200 remittance down to 3% of the transfer amount by 2030, as part of a broader financial inclusion agenda. The global average has remained well above that target, closer to double it, showing how difficult last-mile cost reduction has proven in practice.

Why do real-time payment system linkages like UPI-PayNow matter?

They let a payment settle directly between two countries' domestic real-time payment rails, bypassing the multi-hop correspondent banking chain entirely. This can cut both cost and settlement time dramatically, but it only works for the specific pair of countries that have built the linkage — it doesn't apply to an arbitrary cross-border corridor.

How can a business reduce cross-border payment costs?

Compare total cost (fee plus FX spread against the mid-market rate) rather than the advertised fee alone, evaluate providers corridor by corridor rather than as a single blanket choice, and prefer providers with pre-funded local liquidity in the destination market for recurring payment flows like payroll or vendor payments.

Is the cross-border payment cost problem mostly technical or regulatory?

It's primarily regulatory and structural. Licensing requirements, jurisdiction-specific compliance obligations, and fragmented local payout infrastructure are the main cost drivers — not a lack of faster messaging or settlement technology, most of which already exists.

Teams building or evaluating cross-border payment flows and looking for hands-on help navigating corridor selection, provider comparison, or compliance requirements can reach out to Woyce Technologies.