When a regulator describes a market, the description is only as good as the numbers someone handed over. If nobody collected those numbers, the regulator does not get to skip the exercise. It writes the description anyway, out of whatever is lying around. There is a European example of exactly that, written under a statutory mandate, and it is worth reading before this category finds itself on the receiving end of one.
An assessment fell due before anything had been counted
Article 142(1) of Regulation (EU) 2023/1114, the Markets in Crypto-Assets Regulation, told the European Commission to report to the European Parliament and the Council by 30 December 2024, after consulting the European Banking Authority and the European Securities and Markets Authority, on developments in crypto-assets and in particular on matters the Regulation does not address. Article 142(2)(a) is specific: the report has to assess the development of decentralised finance in markets in crypto-assets and whether regulating it is necessary and feasible.
The two authorities delivered their contribution as a joint report carrying the references EBA/Rep/2025/01 and ESMA75-453128700-1391, dated 13 January 2025 and published by the EBA on 16 January 2025. Paragraph 15 of that report explains what the authorities ran into. On-chain transactions are publicly visible, but the protocols are anonymous or pseudonymous, they can be reached from anywhere, and users can mask their location. So the authorities fell back on anecdotal evidence, meaning surveys and news articles, and on a set of proxies. Paragraph 16 adds inconsistent data, unclear methodologies at the data sources, and difficulty separating decentralised finance from crypto markets in general.
The permission to approximate came with the commission
The most instructive page is not in the analysis. It is Annex 1, which reproduces the Commission's own call for contributions to the two authorities, dated 9 February 2024. On how much European consumers and businesses engage with decentralised finance, that letter says the authorities "may also make assessments based on third party data sources, including approximate ones", and gives the reason in the same sentence: the pseudonymous nature of the transactions and the general data gaps in the market.
That matters for how the rest of this reads. The proxy method was not a shortcut the authorities took and hoped nobody would notice. It was pre-cleared by the institution that would receive the answer, eleven months before the answer was due. Paragraph 17 closes the circle for crypto lending, borrowing and staking: absent reporting obligations on providers, indications of size and participant numbers are limited, and national competent authorities had no concrete statistics on activity in their own jurisdictions apart from registrations held for anti-money-laundering supervision.
Five families of proxy, and the question each one answers
Paragraph 34 of the report names the five proxy indicators it used, and Annex 2, titled Proxies for EU DeFi adoption, works through them. None of them measures the thing being assessed:
- Crypto app downloads. Monthly downloads per 100,000 inhabitants, between 400 and 1,000 in the EU from January 2021 to May 2022, with MetaMask downloads running from 20 to 180 per 100,000 from January 2021 to May 2024, attributed by the authorities to Sensor Tower, Statista and work by Auer and others from 2022. A download is an intention, not a trade.
- Google search trends. Interest in the term DEX over time, and interest across Member States in the names Uniswap, AAVE, Sushiswap and MakerDAO, with values taken as of September 2024. A search is curiosity, including that of people who will never open a position.
- Adoption indexes. The 2023 Global Crypto Adoption Index published by Chainalysis, plus a customised version the EBA and ESMA built themselves from the two sub-indexes that bear on decentralised finance. A score between zero and one is a ranking, not a quantity.
- Euro-denominated transaction volumes. Weekly volume traded against the euro between EUR 1.4 billion in January 2023 and EUR 7.9 billion in March 2024, averaging 8 per cent of volume against official currencies, against 44 per cent for the US dollar and 37 per cent for the Korean won, attributed by the report to Kaiko. The report flags that this speaks to crypto activity in general, not to decentralised finance.
- Time zone estimates. Assigning each transaction to the Americas, to Europe and Africa, or to Asia by the hour of day it happened, producing average shares of 25, 30 and 45 per cent. The report says itself that this is weak for assets traded around the clock from anywhere.
Where the servers sit is not who is trading
Annex 2 adds a sixth measure the five-item list leaves out, and it is the clearest illustration of the problem. The authorities looked at where Ethereum nodes are. Of the 6,245 nodes active in September 2024, 14.78 per cent were in Germany, 4.69 per cent in France and 4.34 per cent in Finland. Then the report says the quiet part out loud: node distribution mostly tracks where the technical infrastructure sits, which makes the count an imprecise proxy for engagement by users.
The euro stablecoin measure gets the same treatment. Dollar denominated stablecoins account for close to 90 per cent of stablecoin market capitalisation and more than 70 per cent of crypto-asset volume traded, euro ones remain marginal, and the report concludes that euro stablecoin use probably understates European engagement rather than measuring it. Each proxy arrives with the direction of its own error named and the size of it unknown.
A finding about an entire segment rests on that ladder
Paragraph 35 then does what reports have to do. It concludes. In aggregate, the authorities write, the proxy evidence points to very low levels of decentralised finance activity in the EU, higher than the world average but below each of the eight comparators it names, among them the United States, the United Kingdom, China and South Korea. Paragraph 36 supplies the one number a reader will remember: the data provider Statista estimates 54 million users of decentralised finance worldwide and 7.2 million in the EU, which the report converts to 1.6 per cent of EU citizens.
Underneath that sit national surveys of something adjacent. The French trade association ADAN extrapolated that 6.5 million people, or 12 per cent of the French population, held crypto-assets in 2024, up from 9.6 per cent in 2023, with 21 per cent familiar with the concept of decentralised finance without necessarily using it. A survey by the National Bank of Slovakia found 6.5 per cent of Slovak citizens holding crypto-assets in 2023, 16 per cent of those holders engaging in decentralised finance, and only half of those doing so regularly. Both are careful pieces of work, and both are surveys about crypto ownership from which engagement in one corner of it is inferred.
The caveat does not survive the one-page summary
The report travels with a factsheet on decentralised finance published by ESMA, which is what most people will read. By our own count it runs to about 515 words. The words proxy and proxies do not appear in it. Neither do Google, downloads, Statista or node. It states that decentralised finance remains a niche phenomenon worldwide and in the EU, that it represents 4 per cent of total crypto-asset market capitalisation, and that the number of users in the EU is estimated at 7.2 million, of whom less than 15 per cent seem to engage regularly.
Nothing there is false, and the word estimated is doing honest work. But the method is gone, and with it any way to see that the 7.2 million is a vendor estimate and the regular share was inferred from a Slovak survey of crypto holders. The qualification lives in the annex, thins in the main text and disappears from the summary, while the number travels everywhere.
The Commission's own pilot found exactly one kind of data missing
Box 3 of the joint report describes something this category should know about. In 2022 the Commission launched a pilot on embedded supervision of decentralised finance, led by Promontory IBM Consulting with the Commission, the EBA, ESMA and EIOPA, and finished in mid-2024. It took eight protocols, Aave, Compound, Uniswap, Curve Finance, Nexus Mutual, Unslashed, 1inch and MakerDAO, and mapped supervisory reporting benchmarks from MiFID, MiFIR, anti-money-laundering rules, COREP, FINREP, SFTR, EMIR and Solvency II onto the public data they emit.
The verdict is more positive than the proxy ladder suggests. Apart from information identifying the persons and entities taking part, the mapping showed that public protocol data meets essential data requirements, and a liquidity pool's balance sheet can be reconstructed from the public ledger back to inception. The five remaining obstacles are practical rather than conceptual: documentation of uneven quality, dataset formats that are not standardised, the absence of personal identification for micro-prudential work, price formation complex enough to need scarce expertise, and unknown interconnections with supervised institutions.
Hot wallets hide the one thing a registered venue already holds
The same passage describes why centralised venues break on-chain observation. Exchanges pool many users into single transactions settled from a hot wallet, so an outside observer sees the exchange's aggregate movement and not what any individual did. The report notes that the hidden detail is known to the exchange, and accessible to its supervisors, financial intelligence units and law enforcement, and that supervision based on public data alone could not identify which users are over-leveraged without conventional supervisory work on regulated entities. Its conclusion is one sentence long: "The involvement of regulated exchanges however ensures that verified identities can be associated with trades".
Read that against the proxy ladder and the asymmetry is complete. The reason a European assessment had to lean on app downloads is pseudonymity, and a venue that opens accounts, verifies identity and matches orders in its own book does not have that reason. The proxy method is not a precedent a registered prediction market venue gets to stand behind. It is what you do with markets that cannot be measured from the inside.
The category enters the European file with no quantity attached to it
The Commission's own Article 142 report was still outstanding as of 18 November 2026. What exists is a targeted consultation on the review of MiCA, open from 20 May 2026 until 30 September 2026, run by the Commission's Directorate-General for Financial Stability, Financial Services and Capital Markets Union, whose answers the Commission says will feed the reports it owes under Articles 140 and 142. Section 4.4 of the consultation document, headed Prediction markets and perpetual futures, is where this category appears. Its opening sentence pairs the emergence of prediction markets built on distributed ledgers with the volume of trading in perpetual futures, so the quantity in that sentence belongs to the other product. Questions 69, 70 and 71 ask whether prediction markets present opportunities or risks, whether they belong under MiFID or MiCA, and what substantive requirements should apply to providers. Not one of the three asks for a figure.
That the category had never been described in a European file before is something we worked through when the consultation opened, in our piece on Question 86 and the open question at the end of the document. The measurement gap has since been written onto the file as a risk in its own right: the consultation's list of identified risks includes supervisory monitoring and data-tracking challenges, with difficulty measuring holdings in self-custodial wallets as the worked example.
Question 70 also decides whose numbers become public
Classification is usually discussed as a licensing question. It is also a publication question. Article 76 of the same Regulation (EU) 2023/1114 tells a crypto-asset service provider operating a trading platform what it has to make public. Paragraph 9 requires bid and ask prices and the depth of trading interest at those prices, continuously during trading hours. Paragraph 10 requires the price, volume and time of executed transactions, as close to real time as is technically possible. Paragraph 11 requires that information to be free of charge fifteen minutes after publication, in a machine-readable format, and to stay published for at least two years, while paragraph 15 keeps order-level data at the competent authority's disposal for five years.
So an answer of MiCA to Question 70 is also an answer that a venue's trade data becomes a published, machine-readable, two-year archive. An answer of MiFID routes the question into the transparency rules that come with that framework instead. Either way, the classification debate settles the measurement question as a side effect, and it is being settled while nobody in the room has a reliable number in front of them.
What makes a figure usable to someone writing under a deadline
Which numbers a venue should publish is a separate argument, and we have made our case for one of them in the piece on open interest and the record volume month. The lesson from the European exercise is about form rather than choice. The five obstacles the pilot listed read, turned around, as a specification for a figure a supervisor can use: the method documented next to the number, period boundaries that do not move between publications, the same denominator every time so two releases can be compared, the basis for assigning activity to a jurisdiction stated rather than implied, and a machine-readable file from which a reader can recompute the headline. The factsheet adds a sixth. The caveat has to be short enough to travel with the number into the summary.
That is Directive 06 in its least glamorous form. A category expecting revenue from many informed participants over years has to be countable over years, by someone outside it, without asking permission. Directive 02 is the other half, because a figure nobody can recompute is not transparency.
A description you do not supply gets supplied without you
None of this is a complaint about the EBA and ESMA. They were handed a statutory assessment of a market with no reporting obligations attached to it, were told in writing that approximate third party sources were acceptable, and did the work with their limitations stated in their own text. The failure sits earlier, in the fact that by the time an assessment was legally due, nothing existed to assess it with.
Prediction markets are roughly where decentralised finance stood in February 2024. A European process has started, it will produce a description of the category, and that description will be assembled from whatever is available when the deadline arrives. Venues hold the raw material the DeFi exercise lacked, because they have accounts, verified identities and their own order books. What they do not yet have is the habit of publishing from it in a form a stranger can use. Firms that would rather be described by their own documented figures than by app store downloads can sign the commitment and start publishing to a standard before one is written for them.