Methodology

Numerator

USD stablecoin supply

Numerator source

DeFiLlama

Denominator

U.S. M2, seasonally adjusted

Denominator source

FRED M2SL

Published as

Percentage, 2 decimals

Plain-English form

1 in N dollars

The percentage is the benchmark. “1 in N U.S. dollars” is a rounded translation of that percentage for readability, derived from the same rounded figure rather than recalculated from raw data, so the two readings can never disagree. A raw, unrounded 1-in-N value is never used as a headline figure.

Companion comparisons may use other denominators. They are labelled separately on the Metrics register and never replace the headline benchmark.

U.S. M2 is a public, independently published measure of the U.S. dollar money stock: cash, checking deposits, savings deposits and small time deposits. The Federal Reserve publishes it monthly as the FRED series M2SL, seasonally adjusted. It is used because it is the broadest widely-cited dollar aggregate that anyone can retrieve and check without a commercial licence.

The reference month is part of the reading. M2SL describes a calendar month, not a day. Every DDD figure is therefore a comparison between a supply observation taken on one date and an M2 value that describes an earlier month. Each frozen Daily Tape record stores its own M2 reference month and states it on the dated permalink; the live reading uses the most recent published observation available at the time it is served.

The release lag. A month’s M2 observation is not available at the end of that month. Across the frozen archive so far, a new reference month has first appeared in a record roughly three to four weeks after the month it describes ended. Between releases the numerator keeps moving and the denominator does not, so DDD drifts upward or downward on supply alone until the next M2 observation steps the denominator.

Vintage limits. Stable Tape stores the M2 value it used, not the full FRED release history. It does not currently retrieve vintage M2SL series, so it cannot show what a past month’s M2 was believed to be at the time versus what it was later revised to. Frozen records preserve the value that was actually used; they are not a substitute for a vintage database. A genuine FRED release and vintage data contract is a precondition for any such claim, and none is in place.

Freshness is not fetch time. These are three different claims and Stable Tape keeps them apart. Observation time is when the upstream source itself last measured the value. Retrieval time is when Stable Tape last read it. Cache window is how long a served response may be reused: the public benchmark endpoints carry a five-minute edge cache. A five-minute cache does not mean supply is measured every five minutes, and a page loaded a second ago can still be showing an observation made much earlier. Where a surface states a time, it states which of the three it means.

The numerator starts from DeFiLlama’s stablecoin universe and narrows to the USD-pegged subset. Each layer below is either included or excluded, with no discretionary adjustment in between.

Official daily cutoff

12:00 UTC

Record window

Snapshot to snapshot

Numerator cadence

Continuous upstream

Denominator cadence

Monthly

The 12:00 UTC snapshot is the official daily cutoff. A Daily Tape record compares the supply captured at that cutoff with the supply captured at the previous one, and stores both snapshot identifiers so the comparison can be reproduced.

“24h” is a label, not a clock. Public daily figures are labelled 24h for readability. Because the cutoff is a fixed snapshot rather than a rolling twenty-four-hour window, the real interval between two consecutive snapshots is sometimes a few minutes to an hour away from exactly 24 hours.

The live benchmark reading and the frozen daily record answer different questions. Use the live reading for current context. Use the dated permalink when a figure needs to stay the same the next time someone opens it.

Each canonical record stores the supply level at the cutoff, the movement since the previous record, the issuer, chain and currency breakdowns behind that movement, the benchmark reading, and the Stablecoin Signal selected from those figures. It is written once and is not overwritten.

Movement is the observed difference between two snapshots. It is not a transaction-level mint and burn feed, and it does not claim to know when within the window a change occurred.

Canonical, bootstrap and gap. The captures dated 17 to 19 June 2026 are bootstrap records from Stable Tape’s initialisation. They stay published and clearly tagged for audit, and are excluded from the canonical series, official highlights, streaks and default citations. 20 June 2026 is the first canonical record, and the citable series begins there. Where a noon snapshot pair was unavailable, the date is published as a declared gap rather than closed up, estimated or backfilled.

Reconstructed history is not a frozen record. Longer historical series shown elsewhere on the site are reconstructed after the fact from published monthly data. They describe the past, but they were not observed and frozen at the time, and they carry no record hash. Only a dated Daily Tape record is vintage-perfect: written at its own cutoff, from the values available at that cutoff, and never rewritten since. The two must not be cited interchangeably.

Hash algorithm

SHA-256

Canonicalisation

stable-json-v1

Digest length

64 hex characters

Permanent address

stabletape.com/tape/YYYY-MM-DD

Each canonical record is serialised through a fixed canonicalisation and hashed with SHA-256. The digest, the algorithm, the canonicalisation and the time it was computed are all stored with the record and shown on its dated permalink.

What the hash proves. It attests that the stored bytes of that record are unchanged since it was frozen. Recompute it from the published record and it will match, or the record has changed.

What it does not prove. It says nothing about whether the source figure was correct. It is not a signature, not a third-party notarisation, and not a timestamp anyone but Stable Tape attests to. External timestamping or a public transparency log would be needed for that, and neither is in place. The archive deliberately keeps the two claims apart: correctness of the input and immutability of the record are different questions.

The highest-priority eligible family becomes the Signal. Eligibility includes the thresholds, cooldowns, repetition controls, attribution checks and data-quality guards below; meeting one numerical condition does not bypass them.

1. Record validity. Incomplete, bootstrap and non-canonical records are flagged before any market reading is attempted.
2. DDD benchmark or milestone. A stable rounded “1 in N” change or a raw DDD move of at least 0.01 percentage points may lead. An immediate one-record reversal in the rounded “1 in N” reading is suppressed when the raw DDD move is at or below 0.012 percentage points. A new U.S. M2 observation leads only when the M2 value changed and absolute net supply movement is at or below 0.05% of supply; routine updates remain evidence.
3. Archive extreme. A record-level daily extreme requires at least 30 canonical records. Until the archive reaches 90 records, the headline states its size.
4. Sequence reversal. The current contiguous run is compared with the immediately preceding opposite run. Both compared legs must be at least $250M, the rounded reversal share must be at least 65%, and comparisons above 200% do not use reversal framing. An eligible partial reversal below 100% states the rounded percentage. A raw share below 100% that rounds to 100% says “almost all” and states the prior amount. Exactly 100% says “fully reversed” and states the prior amount. Above 100% says “more than reversed” and states both amounts. The same bands and amount rules apply in both directions. A one-day move of at least $250M may instead interrupt or end an opposite run of at least three contiguous records; the quantified comparison is evaluated first when both descriptions qualify.
5. Weekly extreme. The largest day in a complete seven-record window must clear $250M and a three-record cooldown, unless it is at least 1.5 times the previous weekly extreme.
6. Large daily movement. Net supply movement of at least $1B becomes the day’s large-move reading.
7. Multi-day streak. Expansion or contraction streaks publish at day 3, 5, 7, 10, 14, 21 or 30, or when an established archive streak is first exceeded.
8. Weekly supply trend. The latest seven eligible canonical records provide directional rolling context when their endpoint supply move is at least $250M. The records need not be date-contiguous.
9. Quiet market with weekly context. A day with absolute net movement at or below 0.05% of supply may carry that latest seven-record trend, unless material offsetting movement takes precedence.
10. Issuer or chain trend. Both endpoints must provide available level data, and the named level must be matched at both endpoints. The leading endpoint change must be at least $250M and at least 50% of total absolute movement on that axis. The aggregate direction must agree with the named move, and the chain bucket “other” is excluded.
11. Offsetting movement. Requires at least $1B of gross expansion plus contraction, absolute net movement at or below 0.03% of supply, a smaller side of at least 0.30% of supply, and a smaller side at least 3 times the absolute net movement. Names appear only when both matched leaders dominate their respective sides.
12. Broad issuer movement. At least 7 of the current top 10 issuers must move with the net direction. Absolute net movement and combined same-direction issuer movement must each be at least 0.05% of supply, and no one issuer may exceed 60% of that same-direction movement. All 10 issuers require matched prior levels.
13. Quiet reading. Absolute net movement at or below 0.05% of supply becomes the fallback only after stronger families and offsetting movement are tested.
14. Net supply movement. A plain directional reading is the final canonical market fallback.

Repetition guards. One prior directional run can produce only one reversal Signal. A different reversal inside the next three canonical records is evidence-only unless its current leg is at least twice the previous reversal leg. Suppressed reversal context is retained where the selected family’s evidence template supports it; rolling supply trends retain that context in the record’s Signal metrics instead of appending a third comparison to the public evidence line. Missing dates and flat records break directional runs.

Guarded attribution: when a mover is not named. One-day issuer and chain moves are clues, not the story by themselves. An issuer, chain or currency is named in the headline only when its move is large enough in dollar terms, accounts for a meaningful share of the day’s total move, and is matched at both endpoints of the comparison. A name is withheld when the prior level is missing, when a name appears or disappears entirely between snapshots — which usually indicates a classification change rather than a real flow — when the aggregate direction disagrees with the named move, or when the comparison is otherwise ambiguous. The residual chain bucket “other” is never named as a mover, because it is a remainder rather than a network. When no single mover can be named with confidence, the record says so instead of promoting the largest available name.

Reconciliation and the residual. Issuer, chain and currency breakdowns are decompositions of the same total, not independent measurements. They are expected to reconcile to the aggregate; where a per-chain decomposition does not account for the whole, the remainder is carried explicitly as “other” rather than distributed across named chains.

History wording. Rolling context uses the latest seven eligible canonical records: the current record and six prior canonical records. Missing dates are not filled and are never treated as zero. “Of the week”, “seven days” and “over 7D” appear only when all seven dates are contiguous. Otherwise the Signal uses gap-safe wording such as “across the last 7 canonical records” or “seven-record endpoint”. Archive-superlative headlines require at least 30 canonical records; below 90 records, the headline itself states the archive size.

Provisional constants. The dollar floors, the 25% evidence-mention floor, the 200% reversal cap and the escalation multiples are provisional. With fewer than 90 canonical records they cannot yet be calibrated against deep history, and they will be reviewed as the archive grows. Frozen records are never rewritten: selection-rule and presentation-template changes apply only to future records, with both versions stored separately on each new Signal.

Sub-signals. Alongside the primary Signal, the Tape and the dated permalink show supporting sub-signals: the DDD reading, the supply movement, and — when a driver can be confidently named — which issuer, chain or currency stood out. When no single mover is confidently named, a sub-signal says so instead.

Drafts and distribution. Stable Tape also generates operator-facing research candidates from the same frozen record, evaluating DDD movement, net and offsetting supply movement, issuer, chain and asset changes, reversals, streaks, historical rank, new tracked listings, leaderboard changes and external money-supply milestones. Stablecoin Velocity is evaluated as research too, and can never select the Signal. The published post for a canonical day always carries the frozen Signal verbatim — the same headline shown on the dated record — with its evidence line and dated permalink. A draft that does not match the frozen Signal word for word is blocked from posting.

Signal selection does not use price, sentiment, trading volume, transaction volume, payments activity, settlement timing, sponsor input, predictions, advice or manual editorial preference.

Production Signal v5, documented above, is the authoritative public methodology. Every Signal on every public surface today is selected by v5 rules and phrased by v5 templates.

A candidate successor, Signal v6, exists as operator-only research. It is a deterministic replay over the same frozen canonical records, exploring a different way of describing a day: whether an observation is a new condition, a materially strengthened existing condition, a state change, a continuation or no new Signal.

What it is not. It is not live. It does not select, phrase or influence any published Signal. It creates no field on any record, publishes nothing, and alters no frozen byte. No candidate result appears on a public surface, and the fact that the engine substantially exists is not a reason to show any of it.

What has to happen first. Calibration against the archive, a methodology freeze, versioning carried in the record, and a separate implementation review for the schema, the website, the cards and the publishing path. Until those are complete and this page is updated to describe a frozen v6 methodology, v5 remains the only public Signal methodology.

Structure layers decompose the same DeFiLlama supply figure the benchmark uses. They are derived, not independently sourced, so they always sum to the total the benchmark divides by M2.

Issuer and chain classification. Which stablecoin belongs to which issuer, and which supply sits on which chain, comes from the upstream pegged-asset dataset rather than from manual assignment. When an upstream classification changes, the shares change with it on the next refresh. That is also why a name that appears or vanishes between two snapshots is treated as a possible classification artifact rather than a real flow.

Velocity is not part of the benchmark. Stablecoin Velocity divides provider-adjusted stablecoin transfer volume by the average USD stablecoin supply over the matching frozen-record window. It is a usage-context research beta: it does not feed the DDD formula, does not enter the frozen Daily Tape, cannot select a Stablecoin Signal, and is not a trading signal. It measures turnover, not active users, payment adoption or investment demand.

• DDD is not financial advice. It is a benchmark and a market-context layer, not a recommendation, a forecast or a trading signal.
• DDD does not include CBDCs. Central bank digital currencies are not stablecoins and are not in the numerator.
• DDD does not include bank deposits. Tokenised deposits, bank-issued tokens and fiat bank balances are not in the numerator.
• The numerator is USD-pegged supply only. Non-USD stablecoins are tracked as currency context and excluded from the headline numerator.
• DDD is not transaction volume. It measures circulating supply, not value moved.
• DDD is not a user count. It does not measure active wallets, unique users or payment adoption.
• DDD is not an endorsement. It is not a risk rating, a solvency rating, a reserve audit or an approval of any issuer, chain, wallet, sponsor or stablecoin. Appearing in the benchmark implies nothing about quality, safety or regulatory standing.
• Stablecoins are not part of U.S. M2. The comparison shows relative scale, not monetary equivalence.
• The two inputs move on different clocks. Supply updates continuously upstream; M2 is monthly and arrives weeks after the month it describes. A change in DDD can come from supply, from M2, or from both.
• The numerator depends on one upstream dataset. A shift in upstream pegged-asset classification changes the tracked total, and a small amount of non-USD-pegged supply can enter it.
• The archive is young. With fewer than 90 canonical records, superlatives are bounded and several Signal constants remain provisional.

A frozen record is never rewritten. When FRED revises M2SL, or the upstream supply dataset restates or reclassifies history, the records that already used the earlier values keep those values. A dated permalink shows what was known at its cutoff, which is the only thing that makes it citable.

So the archive and a live source can disagree, and that is intended. Recomputing today’s figure for a past date is a reconstruction. Reading the record for that date is a vintage observation. Both are legitimate; they answer different questions, and the site does not silently substitute one for the other.

Where a correction is warranted, it is additive. A data-quality note is stored alongside the affected record and printed with it, rather than the stored figures being edited. Methodology changes are described here and take effect from the date of the change forward; they are not applied retroactively to records already frozen.

To cite a specific reading, use the dated permalink: stabletape.com/tape/YYYY-MM-DD. Each dated record is frozen, linkable and shareable, and carries its own dated share card. To cite the benchmark itself rather than one day’s value, cite this page.

Both inputs are public, so the calculation is reproducible: take the tracked USD stablecoin supply from DeFiLlama, take M2SL from FRED for the reference month stated on the record, and divide. The stored snapshot identifiers on each record identify exactly which two observations the movement figure compares.

No sponsor sees an unpublished reading in advance, and no sponsor can veto a negative Signal. Sponsorship never appears beside the benchmark or the Signal, inside dated record content, or in a citation block.

Selection is mechanical and the inputs are public, so this is checkable rather than merely asserted: the rules are published above, and both source datasets can be retrieved by anyone.

What is stablecoin supply vs M2?

DDD divides USD-pegged stablecoin supply by U.S. M2 broad money to show stablecoins as a share of U.S. money supply.

What is stablecoin market cap vs M2?

In this benchmark, stablecoin market cap refers to tracked USD stablecoin supply from DeFiLlama compared with U.S. M2.

How much of U.S. M2 is represented by stablecoins?

The DDD percentage is the answer: USD stablecoin supply divided by the latest available FRED M2SL value.

Does DDD measure stablecoin adoption?

Not by itself. DDD measures monetary scale: circulating USD stablecoin supply relative to U.S. M2. Stablecoin adoption also requires usage metrics such as adjusted transfer volume, velocity, active entities, and payment-like transaction activity. See Stablecoin Adoption Metrics.

Why use U.S. M2?

U.S. M2 is a public macro benchmark for the U.S. dollar money stock, published monthly by the Federal Reserve as FRED M2SL. Stablecoins are global on-chain dollars, not a component of M2 — the comparison is useful for gauging relative scale, but should not be interpreted as stablecoins being part of M2.