How Much of Your Transfer's Cost Can You Actually See?

This matters because almost every consumer protection in this market — comparison tables, fee disclosures, the "no hidden fees" badge — is built around the number that stops being the main number somewhere around the price of a pair of shoes.

If you want the underlying concepts first, read exchange rate markup explained and hidden fees in international transfers. This page is not an explainer. It is a measurement.

The World Bank's Remittance Prices Worldwide database defines the total cost of a transfer as the transaction fee plus the margin taken on the exchange rate. It collects both components separately. It publishes their sum. What it does not publish — and what the UN's SDG 10.c target of "less than 3 per cent" is silent on — is the ratio between them: how much of what a sender pays was legible to them before they pressed send.

That ratio is the whole ballgame for a consumer. A 4% transfer where 4% is printed on the screen is an expensive product honestly sold. A 4% transfer where 0.2% is printed on the screen is something else. The World Bank's 6.49% global average treats the two identically, because as a development statistic it should — the household loses the same money either way. As a shopping statistic it conceals the only variable the sender can act on.

We hold the archive to answer it: 90+ providers, live quotes refreshed every 6 hours, retained daily since March. The missing piece was a defensible way to split one number into two.

The split came from an unlikely place: a tool built for payments operators rather than consumers. Faisal Khan & Co's remittance cost calculator is a modelling instrument for people designing a money transfer business — it asks you to enumerate sending-side charges, correspondent banking deductions on SWIFT routes, and payout partner costs, then tells you what lands.

Its value to us is a single stated rule about the arithmetic:

"Charges come off the send side; only what is left gets converted."

That is a falsifiable claim about how transfers actually work, and it is the hinge of this study, so we tested it before building anything on it. Taking every provider quoting USD to INR on a single day and predicting the receive amount from the fee and the rate alone:

The model reproduces reality to the cent. So we can run it backwards. If receive = (send − fee) × rate, and we independently know the mid-market rate on that day, then every quote splits cleanly into a part the sender was shown and a part they were not:

• True cost = 1 − (received ÷ what mid-market would have delivered)
• Visible = the disclosed fee, as a percentage of the send amount
• Invisible = the margin between the provider's rate and mid-market

Note what the calculator itself does with the rate: it is an input you type in. The tool models three layers of charges in detail and then asks the operator to supply the exchange rate by hand. That is not an oversight — it is an accurate reflection of the industry. The margin is the one cost component that has no schedule to look it up in.

Across all 74,044 observations, weighted by cost: 66.5% visible, 33.5% invisible. The median transfer cost 2.44% in total, of which 0.90 percentage points — about $9 per $1,000 — was margin.

Two-thirds visible sounds tolerable. It is flattered by our measurement point. Our historical archive prices at $100, and $100 is the single most favourable amount at which you could possibly ask this question, because a flat fee looks enormous next to it. A $1.99 fee is 2% of $100 and 0.04% of $5,000. The margin does not move.

So the honest version of the headline is: two-thirds visible on the smallest transfers anyone makes, and falling from there.

The zero-fee cohort

31.1% of all quotes we collected advertised no fee at all. Their median true cost was 1.23% — genuinely cheap, and a real achievement for the category. But the distribution has a tail that the words "no fees" cannot survive:

And the trap works in the other direction too. On the 6,696 corridor-days where a zero-fee option was available, a fee-charging provider was actually cheaper 20.7% of the time. One time in five, paying a fee was the right call.

This is the result we did not expect to be so clean.

To isolate the effect of amount from everything else, we restricted to 241 quote series where the same provider priced the same corridor from the same source on the same day at more than one send amount, and where the fee was genuinely flat across those amounts. Same product, same moment, only the amount differs:

Multiply the transfer by ten and the share of its cost that is visible to the sender falls from 63% to 18%. Nothing about the product changed. The fee did exactly what a fixed fee does, and the margin did exactly what a proportional charge does. Every disclosure regime built around fees is therefore weakest precisely where the money is largest.

There is good news buried in that table: the total cost fell too, from 3.16% to 1.11%. Larger transfers are cheaper in percentage terms. But the composition of what remains is almost entirely the part nobody shows you.

Where each provider crosses over

Because fees are fixed and margins proportional, every provider has a send amount at which the invisible cost overtakes the visible one. Below it, the fee is the main event and a fee-based comparison serves you well. Above it, the fee is a rounding error and the rate is the whole decision.

Two things stand out. Most mainstream MTOs cross over between $113 and $260, at a median of $182 — that is, below almost every real remittance. And Wise effectively never crosses over, because at a 0.02% median margin there is almost nothing on the invisible side to overtake the fee. That is the structural meaning of its pricing model, and it is consistent with what we have measured elsewhere: on major corridors, Wise's markup is genuinely ~0%.

Then there is the group with no crossover point at all, because the margin is not the larger part of their cost — it is the entire cost:

SDG target 10.c commits the world to reducing remittance transaction costs to "less than 3 per cent" by 2030, and to eliminating corridors costing more than 5 per cent. Both limbs are defined on total cost, margin included. But total cost is not what a sender can check, so we scored our archive twice — once on true cost, once on the disclosed fee alone, as a shopper would:

That 20-point gap is the cost of measuring the wrong thing. It is also a caution for anyone reading provider marketing that cites the 3% target: the claim is usually true of the fee and untested on the total.

The counterintuitive part: expensive and opaque are different problems

We expected the corridors failing the 5% limb to be the murkiest. They are the opposite. Of 130 corridors with enough data to judge, 13 had a median cost above 5% — and on almost all of them the cost was openly charged as a fee:

High cost and hidden cost are separate failures with separate remedies. The Gulf and intra-Asia corridors that breach the UN's 5% ceiling are largely transparent about it: the money is taken as a stated fee on a small transfer. Meanwhile the providers with the cleanest "no fees" marketing are, by construction, the ones disclosing least. Transparency regulation and affordability regulation are aimed at different targets, and a policy that only mandates fee disclosure will register no improvement on either.

The practical consequence is measurable, so here it is measured. Across 8,934 corridor-days where at least three providers competed, we asked how often the provider with the lowest advertised fee was also the cheapest transfer:

Sorting by fee gets you the right answer three times in four. That is good enough to feel reliable and bad enough to cost you real money, which is the most dangerous combination a heuristic can have. So:

• Under about $110, comparing fees is mostly fine. That is below the lowest crossover point we measured on any mainstream provider, so the fee genuinely is the dominant cost. Do not over-think a $50 transfer.
• Above about $260, ignore the fee and compare the rate. That is past the highest mainstream crossover we measured, so every provider in the table is taking more from the rate than from the fee. Or better, compare only the number that survives both: the amount that actually arrives. Every provider's fee schedule is a distraction at that size.
• Treat "no fees" as a statement about structure, not price. It tells you where the cost is, not how much it is. Nearly a third of our quotes were fee-free; one in seven of those still cost more than the UN's 3% target.
• Re-check per transfer, not per provider. Across corridors we tracked for 60+ days, the cheapest provider changed a median of 18 times, with 3 different providers taking the lead. There is no provider you can pick once.

This is the entire reason our comparison sorts on the amount that arrives rather than on fees. It is not a design preference; it is the only ordering that is correct at every transfer size.

Stated plainly, because a study that lists no limits has not looked for them.

• Our panel is not the World Bank's panel. Our median true cost of 2.44% sits far below the RPW global average of about 6.49%, and that gap is composition, not contradiction. We price digital providers on corridors with competition; RPW mystery-shops over 360 corridors, including cash agents and retail banks where costs are far higher, and surveys $200 and $500 rather than the $100 our archive prices. Three differences in the same direction, all of which should make our figure lower. Where the panels do overlap the results agree — we measure HSBC at 3.50% margin on top of a $19.88 fee, and RPW has long found banks to be the most expensive channel by a wide margin.
• The archive prices at $100. This is why the visible share of 66.5% is an upper bound, and why the crossover table is the more useful artifact. The $100/$1,000 comparison is measured on real multi-amount quotes, not extrapolated, but it rests on 241 series rather than the full archive.
• Crossover assumes a flat fee. True for most providers we could test, false for those charging a percentage. Those are excluded rather than modelled.
• NGN and GHS payouts are excluded entirely (8,028 observations). Their official mid-market benchmark diverges from the rate transfers actually clear at, producing negative measured markups. Including them would have flattered our cost figures.
• Margin is measured against mid-market on the day of collection, not at the instant of the quote. Intraday movement adds noise to any single observation; it does not bias a median across 74,044.
• This is a frozen study, not a live index. Every figure describes 13 March – 21 September 2026 and will not update. Our cost index and consistency index are the live equivalents.

Dataset. 74,044 provider-corridor-day observations across 185 corridors and 65 providers, collected 13 March – 21 September 2026 and retained daily. Quotes are gathered from provider APIs and comparison feeds every 6 hours. Cost is measured as the shortfall between what arrived and what the mid-market rate would have delivered, so a single figure captures the fee and the margin together; the two are then separated using the send-side identity described above.

Reproducibility. Every number on this page is produced by a single script in our repository, scripts/research/cost-transparency-study.py, run against the same archive that powers the rest of the site. The exclusions and thresholds it applies are the ones documented here.

External sources.

• Faisal Khan & Co, Remittance Cost Calculator — the send-side cost model this study tests and then inverts, and the source of the three-layer framing (sending side, correspondent banking, payout partner).
• United Nations, Sustainable Development Goal 10 — target 10.c: reduce remittance transaction costs to less than 3 per cent and eliminate corridors above 5 per cent by 2030.
• UN Statistics Division, SDG indicator 10.c.1 metadata — defines the indicator as the total cost of sending $200, expressed as a percentage of the amount sent.
• World Bank, average transaction cost of sending remittances — the Remittance Prices Worldwide indicator used for the ~6.49% global benchmark and the provider-category comparison.

Related reading. Average remittance fees 2026 compares cost levels between World Bank data and live quotes; this study addresses cost composition. Exchange rate markup explained covers the mechanism. Our full method and its limits are documented at how we collect and rank quotes.