Six times. That is how far apart two equally defensible estimates of Bitcoin's on-chain transfer value can land, depending on nothing more than an accounting choice made before the analysis even begins. A new working paper from the Bank for International Settlements, No. 1377, puts a number on a problem that on-chain analysts have quietly argued about for years: whether to count change outputs — the leftover Bitcoin a wallet sends back to itself after a transaction — as part of the economic value moved, or strip them out entirely.
Include the change outputs and the transfer-value figure balloons; exclude them and it shrinks to a fraction of that. The paper's authors found this single methodological fork alone drives up to a sixfold divergence in reported figures for the same underlying set of transactions, which means two research desks looking at identical blockchain data can publish headline numbers that differ by 500% or more without either one being wrong, exactly, just answering a subtly different question.
Realized cap told the same story years ago
The paper's authors point to an earlier, now-familiar example of the same underlying issue: Bitcoin's traditional market capitalization — price multiplied by circulating supply — once ran as high as four times its realized capitalization, the metric that values each coin at the price it last moved on-chain rather than today's spot price. That gap has become one of the more closely watched divergences in on-chain analysis, and recent shifts in realized cap dynamics have themselves become a trading signal in their own right, illustrating exactly the kind of confusion the BIS paper is flagging: two numbers, both technically accurate, telling very different stories about the same asset depending on which one gets quoted in isolation.
Ethereum's classification problem is worse
If Bitcoin's ambiguity comes down to one accounting choice, Ethereum's comes down to sheer scale. The researchers examined roughly 67.5 million active smart contracts on the network and found that about 54 million of them — nearly 80% — could not be reliably classified into any clear functional category at all. That is not a rounding error; it means the large majority of Ethereum's on-chain activity resists the kind of clean labeling that would let an analyst say with confidence what a given transaction was actually for.
Related: Bitcoin On-Chain Data Splits: OG Holders Stir as Realized Cap Turns Positive
The stablecoin comparison sharpens the point further. USDT on Ethereum skews toward DeFi-linked activity — lending, liquidity provision, and the composable protocol interactions Ethereum was built to host — while the same token on TRON leans much more heavily toward plain payments and value storage. Same asset, same ticker, functionally different economies depending on which chain it settles on. Aggregating the two into a single, undifferentiated USDT volume figure, as much of the industry still does, papers over a real behavioral split that the researchers argue deserves separate treatment.
A dataset built to make the point
None of this is armchair speculation. The BIS team built its case on roughly 100 billion on-chain records spanning Bitcoin, Ethereum, and TRON — a scale large enough that the classification failures and methodology-driven swings can't be dismissed as noise from a small or unrepresentative sample. That scale is also what makes the paper's central recommendation land with some weight: on-chain metrics, the authors conclude, should be treated as noisy approximations of economic activity rather than direct, ready-made measurements of it.
Why it matters beyond academia
That conclusion has teeth well past a research paper's footnotes. Institutional allocators increasingly cite on-chain figures — transfer volumes, active-address counts, realized cap — as inputs into investment theses, and rating agencies and sovereign-wealth-adjacent research desks lean on the same class of metrics when assessing digital-asset exposure. If any single number can swing sixfold based on an unstated methodology choice, the practical lesson is one of sourcing discipline: know exactly how a figure was computed before treating it as comparable to another, similarly labeled number from a different provider. On-chain demand signals and whale profit metrics that circulate widely in crypto commentary are both downstream of exactly the kind of measurement choices the BIS paper is warning about, which is precisely why the authors frame their findings as a caution rather than a correction — there is no single right way to count a blockchain's economic activity, only more and less transparent ways of disclosing which choice was made.
