BIS: Bitcoin onchain transfer value estimates can vary 6x by methodology
The study also flagged up to a 4x gap between conventional and realized market cap and pointed to large raw-vs-adjusted stablecoin volume spreads.
Bank for International Settlements researchers say widely used onchain “activity” metrics can materially misstate economic activity, after analyzing 100 billion blockchain records across Bitcoin, Ethereum and Tron. The paper found Bitcoin onchain transfer-value estimates can differ by as much as sixfold depending on how transactions are measured.
Key Takeaways
- Bitcoin onchain transfer-value estimates can vary by as much as sixfold depending on the measurement method, and the finding applies to onchain transfer values rather than exchange trading volume.
- The Bank for International Settlements tied the spread to Bitcoin’s transaction structure, particularly whether change outputs and other transfers back to the sender are treated as “value transferred.”
- Bitcoin’s conventional market capitalization has at times been as much as four times higher than realized capitalization, which values each coin at the price when it last moved onchain.
- Visa’s Onchain Analytics dashboard shows $6.4 trillion in total stablecoin volume over the past 30 days versus $313.1 billion in adjusted volume, with the adjusted series designed to filter out bots and internal exchange operations.
BIS Puts a Number on Onchain “Activity” Error: Bitcoin Transfer Value Can Swing 6x
A Bank for International Settlements study is putting a hard ceiling on a problem most onchain users have felt in practice but rarely quantify. After analyzing 100 billion blockchain records across Bitcoin, Ethereum and Tron, the researchers concluded that common indicators can look more precise than the underlying data supports, and that the resulting “activity” narratives can be highly sensitive to how dashboards define a transfer.
The headline number is Bitcoin transfer value. The BIS researchers found estimates of Bitcoin onchain transfer values can vary by as much as sixfold depending on transaction-measurement methodology. The paper is explicit about what that does and does not cover: it concerns onchain transfer values, not trading volume on crypto exchanges.
That distinction matters because transfer value is often used as a proxy for “flows” and “usage” in market commentary, and it is frequently read as if it were a clean measure of economic throughput. The BIS researchers’ warning is that it is not. “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,” they wrote.
The Bitcoin Mechanics Behind the Gap: Change Outputs and Self-Transfers
The sixfold spread is not framed as a mysterious statistical artifact. The BIS ties it to Bitcoin’s transaction structure and, specifically, to how measurement methods treat change outputs and other transfers back to the sender.
Mechanically, a typical Bitcoin spend often creates multiple outputs. One output is the payment to the recipient, and another output returns unspent funds back to the sender as “change.” Onchain, that change is still an output with value, and depending on the methodology, it can be counted as value “transferred” even though it does not represent funds moving to another economic party.
The BIS describes the gap as reflecting differences between measurement methods, including whether those sender-returning outputs are netted out. If a dashboard counts gross output value, it can inflate transfer value by treating internal bookkeeping as external activity. If it attempts to strip self-transfers and change, it can produce a much lower series that is closer to a “payments to others” concept.
For traders, the practical consequence is that a spike in “BTC transfer value” can be a methodological event as much as a market one, especially when providers change definitions, add heuristics, or shift how they cluster addresses. The packet does not include the specific method names or the time window over which the maximum sixfold spread was observed, which leaves readers with the magnitude but not the exact comparison set.
Market-Cap Signals Aren’t Immune: Conventional vs Realized Cap Diverged by Up to 4x
The BIS paper extends the same critique to market-cap based signals, where the temptation is to treat a single number as the value “at risk” onchain. The researchers found Bitcoin’s conventional market capitalization has at times been as much as four times higher than realized capitalization.
Realized cap is computed by valuing each coin at the price when it last moved onchain, rather than valuing all supply at today’s spot price. Traders often use realized cap as a regime tool and as a proxy for holder cost basis, because it weights dormant coins differently from coins that have recently changed hands.
A fourfold divergence between conventional and realized cap is a reminder that the two measures are answering different questions. Conventional market cap is a mark-to-market snapshot of total supply at current price. Realized cap is closer to a “last onchain reprice” measure, which can lag and compress relative to spot-driven market cap when large portions of supply are inactive.
The BIS finding is presented as an “at times” maximum without dates or the market regimes in which it occurred, so the magnitude is clear but the historical context is not. Still, the direction of the risk is straightforward: using conventional cap as a stand-in for economically active value can overstate the base relative to what has actually repriced onchain.
What to Monitor on Dashboards Now: “Adjusted” Stablecoin Volume and Methodology Disclosures
The BIS paper’s broader point is that measurement choices are not a footnote. They are the product. That is especially visible in stablecoins, where the same asset can behave like a DeFi leg on one chain and a payment rail on another.
The researchers described USDT on Ethereum as more closely linked to DeFi activity, while USDT on Tron was associated more with payment-like and store-of-value purposes. They also cited a stark divergence in smart-contract holdings: the share of USDT held by smart contracts on Ethereum exceeded 20% in 2022, compared with around 1% on Tron. Aggregating USDT activity across blockchains, the BIS argued, can conflate different types of economic activity.
A concrete example of “headline number vs filtered estimate” is visible on Visa’s Onchain Analytics dashboard, which is powered by data from Allium Labs. Visa displays both total and adjusted stablecoin transaction volumes and says its adjusted methodology aims to remove potential distortions from high-frequency trading, bots, bridge routing and internal exchange operations. The dashboard currently shows $6.4 trillion in total stablecoin transaction volume across tracked networks over the past 30 days, compared with $313.1 billion in adjusted volume.
For traders comparing dashboards, the immediate checks are procedural rather than philosophical. First, whether a provider discloses how it treats Bitcoin change outputs and other self-transfers in transfer-value metrics, and whether competing BTC transfer-value series converge or remain widely dispersed. Second, whether stablecoin volume is reported as total or adjusted, and whether the provider states what it filters out. Third, whether USDT indicators are chain-specific, separating Ethereum-linked DeFi behavior from Tron-linked payment-like flows, rather than collapsing them into a single cross-chain total.
My Read: Treat Onchain Metrics as Ranges, Not Point Estimates
The filing-cabinet detail here is that the BIS is not saying onchain data is useless. It is saying the most traded-on charts are often definitions pretending to be measurements, and the paper’s own language is blunt that these indicators should be treated as “noisy approximations rather than direct measures of economic activity.”
The threshold that matters is whether analytics providers start publishing enough methodology detail that two traders looking at “transfer value” are at least talking about the same object. If that disclosure standard tightens and series begin to converge, onchain metrics become more actionable as regime tools rather than narrative fuel, and until then the practical edge is treating them as ranges with explicit assumptions, not point estimates with false precision.
