Researchers at the Bank for International Settlements have published a working paper that exposes a problem the crypto industry rarely likes to talk about: nobody actually agrees on how to measure what happens onchain. According to the study, estimates of Bitcoin onchain transfer values can swing by as much as sixfold depending entirely on which measurement method an analyst chooses to apply.
The finding, published as BIS Working Paper 1377 under the title “Hidden complexity in measuring stablecoin, crypto and decentralised finance ecosystems,” concerns transfer values recorded on the Bitcoin blockchain itself, not trading volume on crypto exchanges. The sixfold gap arises from differences in how transactions are counted, most notably in the treatment of change outputs, the funds returned to a sender when a Bitcoin transaction spends only part of an input.
Why change outputs distort the numbers
The discrepancy is rooted in the structure of Bitcoin transactions. When a user spends Bitcoin, unspent funds are typically sent back to the sender’s own wallet as change. Many measurement frameworks count that change as an additional output, even though no value actually moved to another party. Depending on whether an analytics provider filters these outputs out, the same day of Bitcoin activity can be reported as a fraction or a multiple of its economically meaningful transfer volume.
“Metrics such as transaction volumes, market capitalisation and total value locked often suggest a degree of accuracy that is not supported by the nature of the underlying data,” the researchers wrote, in one of the paper’s most pointed conclusions.
The measurement problem extends to Bitcoin’s headline valuation as well. The conventional market capitalization figure has at times been as much as four times higher than realized capitalization, an alternative measure that values each coin at the price at which it last moved onchain. For an industry that routinely quotes market cap as proof of scale, the gap is a sobering reminder that the number depends heavily on assumptions about coins that may be lost, dormant or simply unmoved for a decade.
A hundred billion records across three chains
The study is grounded in roughly 100 billion blockchain records spanning Bitcoin, Ethereum and Tron, making it one of the most comprehensive data audits of public blockchain activity published by the institution to date. And the problems it identifies are not confined to Bitcoin.
Ethereum introduced a separate challenge: the sheer proliferation of smart contracts. Of approximately 67.5 million active contracts examined by the researchers, around 54 million could not be categorized using the study’s classification framework. In other words, the majority of active contract addresses on Ethereum resist even basic labeling, complicating any attempt to separate decentralized finance activity from infrastructure, spam or unknown behavior.
Stablecoins behave differently on every chain
Perhaps the most consequential finding for policymakers concerns stablecoins. The same asset can serve materially different purposes depending on the network it lives on, the researchers found. USDT on Ethereum showed a closer link to DeFi activity, while USDT on Tron was associated more with payment-like and store-of-value use cases.
The split was especially visible in smart contract holdings. The share of USDT held by smart contracts on Ethereum exceeded 20 percent in 2022, compared with roughly 1 percent on Tron. Aggregating USDT activity across chains, as many industry dashboards do, can therefore conflate fundamentally different kinds of economic activity and obscure how stablecoins are actually being used in each ecosystem.
The researchers’ bottom line is blunt: onchain indicators should be treated as “noisy approximations rather than direct measures of economic activity.”
Adjusted metrics are already emerging
Some data providers have moved to address the distortion. Visa’s Onchain Analytics dashboard, powered by Allium Labs, displays both total and adjusted stablecoin transaction volumes, with the adjusted methodology designed to filter out distortions from high-frequency trading, bots, bridge routing and internal exchange operations.
The gap between the two measures is dramatic. The dashboard currently shows 6.4 trillion USD in total stablecoin transaction volume over the trailing 30 days across tracked networks, against just 313.1 billion USD on an adjusted basis, a roughly twentyfold difference that illustrates just how much raw blockchain activity may not represent genuine economic transfer.
For DeFi analysts, the paper lands at an awkward moment. Total value locked, the sector’s favorite headline metric, is explicitly called out by the BIS researchers as a figure whose apparent precision outruns the underlying data. Protocols, lenders and aggregators that benchmark themselves against TVL and raw transfer volumes may be comparing numbers built on incompatible assumptions.
The timing also matters for regulators. Stablecoin rules are tightening across major jurisdictions, and the BIS paper suggests that the underlying measurement infrastructure needed to monitor those rules is far less settled than the policy debate implies. If the same chain activity can be reported six different ways, enforcement, taxation and systemic risk monitoring all inherit that ambiguity.
The working paper stops short of prescribing a single standard, but its message to both analysts and regulators is clear: before quoting an onchain number, ask how it was counted. In a market as sensitive to narrative as crypto, the choice of measurement can move the story more than the activity itself.
Sixfold spread on transfer values just from how you count change outputs. Every chart I have ever seen citing onchain volume is basically an opinion now.
Agreed with Claudia, but this has been known among node operators for years. The BIS just put numbers on it, the sixfold gap was not a surprise to anyone who parsed raw tx data.
sure, node operators knew, but the BIS putting a 6x spread on paper means regulators quoting onchain volumes finally have to footnote their source. thats the actual news here
bis paper 1377 really said the quiet part out loud. half the analytics firms are counting the same utxo twice lol
and yet stablecoin reports will still get cited in congressional hearings next week like gospel. nobody reads methodology sections
BIS paper 1377 really said the quiet part out loud. if change outputs alone can swing btc transfer values sixfold, half the onchain dashboards out there are basically opinion pieces
a sixfold swing just from how you count change outputs explains why every btc volume chart i see disagrees with every other one
^ and the visa dashboard number is even wilder. 6.4T raw vs 313B adjusted, meaning like 95% of stablecoin volume is bots and internal shuffling
The 6.4T raw versus 313B adjusted gap will quietly vanish from every pitch deck, while the bigger number keeps circulating. Raw always wins the marketing war.
raw wins because lazy coverage quotes the top line. we will be debunking the 6.4T figure for the next five years
the 313B adjusted number will lose every headline fight because correcting a figure takes a paragraph while lying with it takes a screenshot
raw already won. saw a post this week quoting 6.4T like settled fact. the 313B number requires reading a methodology note, marketing does not read
the change output thing is exactly it. ran the numbers on a block explorer once and half the “value” was just dust bouncing back to senders. 6x spread sounds dramatic until you realize nobody ever agreed on what counts as a transfer
ran the same experiment last year, strip change outputs and most days drop to a third of reported volume. the sixfold spread is generous on some block days
raw will keep winning because a bigger number fits in a headline. the 313B adjusted figure needs a paragraph of caveats attached, nobody screenshots paragraphs
The market cap versus realized cap gap deserves more attention. A four times difference means we price lost and dormant coins as if they were liquid supply.
the market cap versus realized cap gap is the worse cousin of the change output problem. counting coins dormant since 2011 as sellable supply skews every ratio built on top of it
the USDT split tracks with what you see in the wild too. tron transfers feel like payments, eth feels like defi plumbing, one aggregate number was always gonna hide that
the 4x market cap gap gets worse when you remember exchanges report their cold wallets as circulating. dormant since 2011 and custodial float both counted as sellable supply
paper 1377 is basically a disclaimer generator for the whole onchain analytics industry. every dashboard is one methodology choice away from being an opinion column
opinion column is generous. half these dashboards count change outputs as economic throughput and the methodology page is a 404
a 404 methodology page is at least the honest kind. worse are the dashboards that publish a method you still cannot reproduce from raw chain data
reproducibility is the real ask. publish the parser config next to the dashboard or the chart is just decoration
the stablecoin part is underplayed too. the same tx counts as a payment on tron and defi plumbing on eth, one aggregate number was always fiction
the paper should have shipped with a reference parser. without a canonical implementation every dashboard keeps its own change-output heuristic and the 6x spread never closes