Stablecoin Volume Is Not Adoption: Classify the Flow Before You Count It

Treasury Desk Brief · Stablecoin Watch

Public-source measurement review Source retrieval: 2026-09-08 Classification + reconciliation

Short executive summary

A stablecoin transfer is an observation. A payment is an economic event. Adoption is a claim about who is using a stablecoin, for what job, how often, across what population and period.

Public blockchain volume can mix payments, trading, treasury moves, mint/burn events, bridge legs, refunds and unknown flows. The practical response is not to dismiss volume. It is to classify the flow before using volume as evidence of a specific kind of adoption.

Measurement rule: “Adjusted volume” is not a portable synonym for “payment volume.” Record the exact dataset, field, version, filters, purpose labels and counting unit.

Start by defining what “adoption” means

For measurement work, “stablecoin adoption” is too broad to stand alone. A useful adoption statement should identify the actor, the job being done, the covered universe and period, and the evidence that shows incidence, repetition or breadth for that job.

For a payment use case, useful evidence can include provider records tied to known payment categories, repeated payment events, merchant or supplier obligations, organisations using the rail over time and reconciliation evidence. For a store-of-value use case, balances, holding periods and holder cohorts can matter more than transfer velocity.

This is an analytical definition, not an industry standard. A high transfer total may be evidence of rail activity without establishing broad adoption of merchant payments, B2B payments or another named use case. The reverse error matters too: trading, financial settlement, treasury management and store-of-value behavior can be genuine uses even when they are not merchant payments. The Bank for International Settlements’ 2026 review discusses trading and offshore store-of-value demand as stablecoin use cases alongside payments.

Three methodologies can answer three different questions

Visa / Allium: adjusted onchain activity

Visa Onchain Analytics publishes raw and adjusted stablecoin transfer metrics. Its methodology uses labels and heuristics, including a largest-transfer-within-transaction rule and address-activity thresholds. But its listed adjusted categories include centralized exchanges, decentralized exchanges, lending, investment funds, minting and burning, ramps and other activity. “Retail sized” is a size bucket for adjusted-category transfers below $250, not proof of retail purchase purpose.

Current Allium stablecoin metric documentation separates adjusted activity from later intent and payment-purpose classification. The current schema says adjusted USD volume can include Real-World Payment, Investment/Trade or Store as Value, while the Payments Pipeline has a later purpose stage and an Unclassified outcome. The exact field, table and version therefore matter.

Artemis: bottom-up known payment activity

Artemis’s Ground-Up methodology starts closer to payment purpose. It aggregated data from 20 payment companies and supplemented estimates for 11 more, for 31 firms in total, focusing on categories such as B2B, B2C, P2P and cards while excluding investment flows.

The trade-off is coverage. Artemis says the sample is not exhaustive, provider overlap can create duplication, some charts rely on subsets, and growth partly reflects firms entering the dataset. Its February 2025 month annualised to $72.3 billion of known stablecoin-based payments at the time of the study. That is a dated sample estimate, not a universal ceiling.

McKinsey / Artemis: modelled payment categories

A February 2026 McKinsey-Artemis analysis used a hybrid approach. Known contracts and tags were used where possible, while harder categories used assumptions: the B2B estimate assumed 20% of tagged custody and treasury activity represented true B2B payments, and the remittance estimate assumed 10% of selected stablecoin-to-fiat trading volume reflected remittances.

The resulting estimate was about $390 billion of annualised stablecoin payments based on December 2025 activity. McKinsey compared that with its own Global Payments Map and reported about 0.02% of global payment volume. It is a model result under stated assumptions, not transaction-level ground truth, and should not be divided by a separate gross-onchain headline to manufacture a universal payment share.

Stablecoin Flow Classification Sheet

Primary bucket	Payment-adoption treatment	What it does not prove
Bridge / wrap / migration	Separate; retain as rail activity	Not fraud; may be linked to a later payment
Mint / burn	Separate supply mechanics	Not an end-user payment by default
Same-entity transfer	Separate when common control is documented	Not wash/fraud by itself
Return / refund	Separate and reconcile under gross/net policy	Does not erase the original gross event
Trading / DeFi / financial use	Separate from a named real-economy payment numerator	Not fake or non-economic
Confirmed payment-related transfer	Include on transfer basis	One transfer leg is not necessarily one economic event
Store-of-value positioning	Separate; use stock/cohort evidence	Low velocity does not prove no use
Unknown	Keep visible and unallocated	Neither payment nor bot/fraud without evidence

Author synthesis. Use one mutually exclusive primary bucket per counting basis; preserve overlapping actor, rail, lifecycle and confidence evidence as tags. Unknown remains visible.

A practical precedence is ingestion duplicate → documented bridge/migration → mint/burn → documented same-entity → linked return/refund → trading/DeFi/financial use → confirmed payment → store-of-value → unknown. Freeze the policy before calculation so one observation is not subtracted twice.

A July 2026 transaction-level study of Austrian CASPs illustrates why entity evidence matters. Using a regulatory registry of addresses controlled by all 12 Austrian CASPs registered at end-2024, the authors reconstructed BTC, ETH, USDC and USDT activity through May 2025 and found 41.1% of total value in their dataset was within-entity bookkeeping. That is evidence for this sample, not a global stablecoin percentage. The same paper explains that public blockchain addresses do not themselves reveal geography or identity and that custodial customer balances can exist in offchain internal ledgers.

Count observations and economic events separately

SYNTHETIC EXAMPLE — NOT MARKET DATA. Every synthetic token is valued at exactly USD 1.00. The dataset supplies explicit event links only for demonstration.

Transfer-basis reconciliation	USD
Raw ingestion	$34,458
Less exact ingestion duplicate	($1,000)
Deduplicated gross rail activity	$33,458
Payment-classified transfer legs	$3,060
Known separate classes	$29,998
Unknown	$400

On the economic-event basis, the two $1,000 transfer legs of the first synthetic payment are explicitly linked to one $1,000 obligation. Adding the four confirmed payment events gives $2,060 gross. Applying the stated same-period policy for an explicit $100 refund gives $1,960 net.

Do not mix units: $3,060 of payment-classified transfer legs and $2,060 of linked payment events are different counting bases. Do not subtract the event total from gross transfer activity as though it were a transfer bucket.

What stronger adoption evidence looks like

For B2B payments, useful evidence can include confirmed payment-event value and count, participating organisations under a documented entity method, repeat usage by cohort, corridor/counterparty coverage and the share of events reaching the required settlement and reconciliation state. A public onchain success flag alone does not prove invoice discharge, beneficiary credit or accounting closure.

For consumer payments, transaction size can be descriptive, but a small amount is not a purpose label. Addresses are not users, an omnibus address can represent many users, and one user can control many addresses. Provider or merchant evidence can therefore be more informative than address counts for end-user adoption.

For financial settlement, trading or liquidity management, use a different adoption definition. Excluding those flows from a real-economy payment numerator does not make them fictitious. For store of value, use balance, duration, concentration and cohort evidence; low velocity is not evidence against holding use.

Finally, keep unknown visible. A residual after filters is not automatically a payment. An uncertain transfer is also not automatically a bot, wash trade or fraud. Classification error can overstate payments or remove genuine ones.

What this does not prove

This framework does not estimate global stablecoin adoption or global payment market share. The synthetic dollar amounts are instructional only. It does not say raw volume is useless, adoption is absent, or excluded activity is fake.

It also does not claim that a public onchain dataset captures offchain custodial transfers, fiat legs, beneficiary credit, contractual discharge or reconciliation closure.

Reader action: before using a stablecoin volume number, write down the observation unit, the purpose-classification rule, the denominator and the unknown bucket. Then reconcile transfer legs to economic events only when the linkage is defensible.

Treasury Desk Brief is educational public-source analysis. It is not investment, legal, tax, accounting or compliance advice.