BIS finds one accounting choice swings Bitcoin's transfer volume sixfold

Start with the change output. Spend 0.5 BTC from a wallet holding 1 BTC and the entire 1 BTC moves onchain, with 0.5 coming back to the sender as change ^[3]. Count the whole movement and you have recorded 1 BTC of transfer where the economic transaction was 0.5, exactly twice the number ^[4].

Read the reported factor the other way and it gets easier to price. If the generous convention is six times the strict one ^[2], the strict one is about 17% of the generous one, an 83% gap between two defensible figures describing the same chain ^[1]. Neither is a mistake, and only one of them is in your spreadsheet.

Ethereum is a filtering problem instead of a change-output one. Tokens are roughly 11% of the active contracts BIS counted, which leaves about 89% of registered contract activity for somebody to classify ^[3]. Every swap through a decentralised exchange involves multiple contract calls, and one user action can trigger a cascade of events that a naive block explorer reads as activity ^[5].

The paper found token issuance and decentralised exchange metrics carry the same distortion risk ^[11]. It also found that stablecoins behave differently by venue: on Ethereum they are embedded in lending protocols, exchange liquidity pools and yield strategies, while on Tron holdings sit predominantly outside smart contracts altogether ^[7].

BIS's own conclusion is that onchain indicators should be treated as rough approximations and not the precise economic metrics many participants assume ^[10]. Crypto Briefing's account of the paper does not break the record count down by chain or say which convention any commercial data provider uses ^[13].

In my view the sixfold range is a labelling problem before it is a pricing problem. Inside one provider's consistent convention, a series still tells a subscriber whether activity rose or fell quarter on quarter, and that is what most allocation work actually consumes. The counter-thesis is that diligence and cross-chain comparison are precisely where conventions differ. BIS makes that point directly: two analytics platforms reporting the same metric for the same blockchain can produce vastly different numbers if they handle unspent outputs or filter contract interactions differently ^[8]. Disclosure settles which reading holds. If the large providers publish their treatment of change outputs and their filtered numbers land within a few percent of each other, the sixfold range stays a laboratory result. If they land three times apart, every fee and valuation model keyed to onchain volume needs restating.