Not a Transaction Fee but a Boundary Toll: Where the Regressivity of Onchain Finance Actually Originates

The Distance Between "Usable" and "Safely Usable"

On August 10, Decrypt published a contributed piece by Pauline Shangett, Chief Strategy Officer at the crypto platform ChangeNOW. Her argument is direct: permissionless access answers only the narrow question of whether a person can enter the system, while financial inclusion requires a separate question about whether they can use it safely.

She introduces the concept of an "error budget," the amount of money any user can afford to lose while learning how a product works. When two people perform the same operation and each pays 25 dollars in fees, the person moving 10,000 dollars experiences an annoyance while the person moving 100 dollars watches a quarter of the transfer disappear. Protocols do not know a user's income or savings, and arguably should not. Products, however, can see the amount being sent, estimate the network fee and calculate the expected amount on arrival. Many interfaces nonetheless present technical decisions as though every user has identical capital, experience and tolerance for loss.

The observation reads as intuitively correct. The question is at what magnitude it actually occurs.

The Measured Data: Cost Accumulates at the Boundary, Not on the Chain

On this point, a research paper published by Banca d'Italia on July 30, 2026 supplies decisive material. Titled "Are Stablecoins Efficient for Remittances?" and issued as number 86 in the bank's Markets, Infrastructures and Payment Systems series, it is the first mystery shopping exercise on stablecoin remittances conducted by a major central bank.

The methodology is rigorous. On March 24 and 26, 2026, researchers executed real transfers of 200 USDC in both directions across ten corridors linking Italy with Argentina, Brazil, South Africa, the United Arab Emirates and Japan. The 200 dollar amount was chosen to match the World Bank's standard for measuring remittance costs. Transactions were processed on Ethereum, and each transfer was decomposed into five stages: funding the exchange account, purchasing USDC, transferring on-chain, selling USDC, and withdrawing local currency.

Total costs ranged from 0.30 percent to 8.96 percent of the amount sent. The most important finding is that the on-chain transfer accounted for only around 0.4 percent on average, with the bulk arising from exchange purchases, funding methods, withdrawals and currency conversion.

Specific cases illustrate the structure. On the UAE-to-Italy corridor, where a credit card was the only available funding method, a 3.8 percent surcharge pushed the total to 8.95 percent. At the other end, the bank cautions against reading the cheapest result, 0.30 percent on Italy to Argentina, as evidence of efficiency: the figure reflects the gap between Argentina's official exchange rate and the rate implicit in USDC transactions, which stood roughly 3 percent above the official benchmark that day.

Against Wise, one of the largest money transfer operators, USDC was cheaper on three corridors and more expensive on four. Brazil to Italy produced the clearest gap, with USDC at 2.21 percent against Wise at 4.68 to 4.89 percent, while both UAE directions and Italy to Brazil favored the incumbent.

Set alongside World Bank statistics, the picture sharpens. The latest Remittance Prices Worldwide edition records a global average cost of 6.36 percent for sending 200 dollars in the third quarter of 2025, with Sub-Saharan Africa the most expensive destination region at 8.46 percent. Digital remittances averaged 4.59 percent and the digital-only operator index 3.54 percent. Stablecoin remittance, then, matches or beats established digital services on good corridors and matches the world's most expensive region's traditional services on bad ones.

Why a Fixed-Cost Structure Is Regressive

It is worth noting that regressivity is not unique to crypto. In the same World Bank data, sending 200 dollars costs 6.36 percent globally while sending 500 dollars costs 4.08 percent. Traditional remittance already charges a higher rate on smaller amounts.

The problem is that onchain technology did not dissolve this structure. The cost of moving value on a blockchain approached zero, but fixed costs did not disappear so much as relocate to the entry and exit points with fiat currency. Exchange funding and withdrawal fees, card surcharges, foreign exchange spreads, local bank crediting charges: each contains a component that does not scale with the amount sent.

The cognitive costs the contributed piece identifies concentrate in the same place. Which network to select, which permission to grant, which route delivers how much at the far end. Every one of these judgments arises at the moment of crossing between the chain and the existing financial system. That monetary and cognitive costs pool at the same boundary is not coincidence. A boundary is where responsibility changes hands, and because it remains unsettled who bears that responsibility, the cost is passed to the user.

Repeated User Error Is Product Data, Not Carelessness

Read against that structure, the piece's central claim carries more weight. One user erring once is an individual judgment, but when the same category of error recurs across different users, the cause probably lies in product design.

The industry's default response is education: read the documentation, watch the tutorial, understand how gas works, learn the difference between a network, a wallet, a bridge and an exchange before sending money. Education matters, but it cannot excuse predictable failure modes. Sending 100 dollars should not require a degree in distributed systems, as the author puts it.

The problem bites hardest where crypto has become an instrument of daily life rather than speculation. According to Chainalysis, Sub-Saharan Africa recorded more than 205 billion dollars in on-chain value received between July 2024 and June 2025, up roughly 52 percent year on year, with Nigeria alone accounting for 92.1 billion. Transfers of 10,000 dollars or less made up more than 8 percent of value received in the region, against a global figure of 6 percent. In March 2025, monthly volume spiked toward 25 billion dollars following a sharp naira devaluation.

For people retreating into dollar-pegged stablecoins amid inflation and foreign exchange scarcity, the error budget is very small indeed. A single misdirected transfer is not a learning cost but lost living expenses.

Business Development Insights

1. Design the cost structure of small-ticket services around boundary-crossing costs, not onchain costs. Banca d'Italia's measurements show the onchain portion averaging around 0.4 percent of total cost. Competitiveness in a remittance product therefore depends not on chain selection but on how far the fiat entry and exit points can be brought in-house and optimized. Domestic corporate stablecoin platforms that quote a flat per-transfer fee are pointing in the right direction, but a flat fee necessarily means the rate spikes on small transfers. For anyone targeting small-ticket users, whether the ramp fee is structured as a percentage or a flat amount is, in effect, the choice of which customers you serve.
2. Measure error rates as a product KPI, not a customer support metric. To operationalize the claim that repeated error is product data, organizations should continuously track network-selection error rates, failure rates from unsupported destinations, divergence between expected and actual received amounts, and post-confirmation reversal requests, and feed those directly into remediation priorities. These are often computable from logs already held. They also serve as the criterion for separating problems that warrant more educational content from problems that require an interface redesign.
3. Internalizing cognitive cost is the differentiating feature of Japan's implementations. Non-custodial wallets in Japan that sponsor gas fees, complete identity verification through public personal authentication and operate via passkeys have reached general availability. This is usually described as a convenience improvement, but its economic significance is larger: the cost of technical judgment previously borne by the user has been absorbed by the operator. That absorption costs money, and who recovers it and how is the core of the business model. Conversely, an operator without a path to recovering the cost of internalization will ultimately push it back onto users. Any organization claiming inclusion needs to design the allocation of cognitive cost alongside its revenue model, not after it.

Sources

• Decrypt, "Web3 Is Building Financial Freedom Only for People Who Can Afford Mistakes" (Pauline Shangett, August 10, 2026) https://decrypt.co/375226/web3-is-building-financial-freedom-only-for-people-who-can-afford-mistakes
• Banca d'Italia, "Are Stablecoins Efficient for Remittances? Evidence from a Mystery Shopping Exercise" (Markets, Infrastructures and Payment Systems, No. 86, July 30, 2026) https://www.bancaditalia.it/media/notizia/are-stablecoins-efficient-for-remittances-evidence-from-a-mystery-shopping-exercise-by-banca-d-italia/
• CoinDesk, "Bank of Italy research suggests stablecoins aren't necessarily cheaper for remittances" (August 1, 2026) https://www.coindesk.com/business/2026/08/01/bank-of-italy-research-suggests-stablecoins-aren-t-necessarily-cheaper-for-remittances
• Payment Expert, "Stablecoin remittances cost up to 8.96%, Bank of Italy finds" (August 3, 2026) https://paymentexpert.com/2026/08/03/bank-of-italy-stablecoin-remittances/
• World Bank, "Remittance Prices Worldwide – Issue 54, September 2025" https://remittanceprices.worldbank.org/sites/default/files/2026-04/RPW_main_report_and_annex_Q325.pdf
• Chainalysis, "Sub-Saharan Africa Shows Strong Crypto Retail Activity" (2025) https://www.chainalysis.com/blog/subsaharan-africa-crypto-adoption-2025/