Payments companies, regulators and analysts should treat headline blockchain metrics as estimates rather than direct measures of economic activity, according to a new working paper published by the Bank for International Settlements.
The paper examines granular records from Bitcoin, Ethereum and Tron and identifies three sources of measurement error: Bitcoin’s transaction architecture, the difficulty of classifying programmable smart contracts, and the different uses of the same stablecoin across networks. The authors analyzed data structured by Mercurius, a research initiative operated by De Nederlandsche Bank and developed with the BIS Innovation Hub and Deutsche Bundesbank. The dataset comprises 100 billion blockchain records.
The findings are those of the paper’s authors and do not necessarily represent the views of the BIS, its member central banks or De Nederlandsche Bank.
Bitcoin transfer estimates vary with the method
Bitcoin transactions consume unspent transaction outputs in full and commonly return the unused portion to the sender as change. A simple aggregation can therefore count technical change outputs alongside value sent to another party.
The researchers found that estimated Bitcoin transfer value varied by as much as a factor of six depending on the methodology used to identify and exclude change. That range concerns onchain transfers, not exchange trading volume. The paper also found that simple market-capitalization measures have at times reached four times realized capitalization, which values coins at the price when they last moved.
For payment businesses, the practical lesson is that a transfer recorded by a ledger is not automatically equivalent to a customer payment or settlement between separate economic parties. Transaction structure and wallet behavior must be interpreted before volume is used for market sizing, compliance monitoring or operational benchmarking.
Token labels do not solve classification
Programmable networks create a different problem. The researchers classified 13 million active Ethereum contracts using common technical standards, while 54 million active contracts remained outside that classification. They identified 1.4 million contracts issuing tokens and found extensive reuse of familiar symbols: approximately 7,000 token contracts used the USDT label.
This does not mean that every unclassified contract is economically irrelevant or that every repeated symbol is fraudulent. It shows that labels and standard event logs alone are insufficient for measuring genuine payment, trading or financial activity. Smart-contract interactions may be distributed across transaction inputs, execution traces, logs and changes in contract state, requiring technical classification and judgment.
The same stablecoin can reflect different activity
The study also cautions against combining stablecoin activity across chains without considering the infrastructure. The authors found that the share of USDT held by smart contracts on Ethereum exceeded 20% at one point, compared with roughly 1% on Tron. They associated Ethereum activity more closely with decentralized-finance services such as liquidity provision and trading, while activity on Tron appeared more connected to smaller transfers, exchange holdings and payment-like or store-of-value uses.
Those patterns do not identify the purpose of every transaction. They do show why a single aggregate for USDT can mix economically different behavior. A payments dashboard that combines all chains may overstate commercial payment use if it includes DeFi movements, internal exchange transfers or technical contract activity without distinction.
What better measurement looks like
The authors recommend complementing point estimates with bounded ranges that reflect protocol-specific uncertainty and can be revised as classifications improve. They also call for explicit assumptions, technical classification and disaggregation by both asset and network.
That approach has direct implications for payment processors, stablecoin issuers and financial institutions. Metrics used in product planning or regulatory reporting should document whether they measure raw ledger transfers, adjusted economic transfers, wallet-to-wallet movements or identified merchant payments. Cross-chain comparisons should apply consistent definitions while preserving differences in network architecture and user behavior.
Public blockchains provide unusually detailed records, but transparency at the ledger level does not remove the need to interpret what those records mean. The paper’s central warning is that precise-looking totals can conceal large methodological uncertainty—a risk that becomes more important as stablecoin data is used for payments oversight and financial-stability analysis.