# How ReserveBeat measures stablecoin reserves: a methodology deep-dive
A walk through the measurement methodology behind the ReserveBeat coverage board: storage-proven on-chain supply, hash-anchored attestation parsing, same-document coverage ratios, and an append-only evidence chain — with the real limits stated beside each figure.
Every figure on the [coverage board](/) was measured under one
reviewed [methodology document](/methodology), and this post walks through
how that measurement actually works — including what it refuses to do.
Nothing here is a solvency claim about any issuer: tracked tokens are
subjects of public data, and no tracked issuer is a customer of ours.
The instinct in this category is breadth: more tokens, more chains, more
tiles. But a dashboard is only as trustworthy as its least-supported
figure, and breadth is usually purchased by relaxing sourcing standards
one tile at a time. ReserveBeat's bias runs the other way — fewer
figures, each carrying its source tier, its as-of date, and its stated
limits, with absence rendered as absence. A reader who knows exactly how
a number was measured can decide what it is worth; a reader facing a
wall of unsourced numbers can only decide whether to trust the wall.
## On-chain supply: proven where possible, labeled where not
Supply is snapshotted at finalized Ethereum blocks through a quorum of
independent RPC providers that must agree byte-for-byte. Each figure
carries a provenance tier. The stronger tier, STORAGE-PROVEN, means the
total-supply storage slot was read via eth_getProof and the proof verified
locally against the block's state root. The board prints the tier beside
every supply figure — see the [source-tier glossary
entry](/learn/source-tier) for the weaker QUORUM ETH-CALL tier.
If quorum legs diverge, the snapshot blocks: the divergent responses land
as evidence, an incident opens, and nothing publishes from a disputed read.
Averaging divergent reads would publish a number that no provider
actually returned. If independent providers disagree about a finalized
block, the disagreement *is* the finding — a provider is wrong, or an
assumption is — and the honest output is an incident, not a midpoint. A
blocked snapshot costs one data point. A smoothed-over divergence costs
the reason to believe every other data point.
## Attestation parsing: capture first, refuse on drift
Issuer reserve reports are captured and hash-anchored **before** parsing.
Parsing runs against pinned per-issuer templates that refuse on layout
drift: a document that does not cross-foot, or whose section anchors moved,
lands nothing beyond the captured source. Line labels are stored verbatim —
normalization never discards what the issuer actually said. Issuer reissues
and parser upgrades are both handled by
[restatement](/learn/restatement): a new parse supersedes the old row,
and the old row stays independently verifiable.
## Coverage ratios: the honesty core
A [coverage ratio](/learn/reserve-coverage-ratio) only ever divides two
figures from the **same** issuer attestation at the **same** as-of date.
Circle's May 2026 USDC report states total reserve assets of
and USDC in
circulation of ,
a ratio of 1.001004 at that date. On-chain measurements never enter a
ratio. USDC's stored ratios inside the public window:
Reports whose stated scope is a consolidated group — one redemption value
across every token an issuer issues — yield **no per-token ratio** unless
the document itself itemizes the token. That is why the board's USDT page
shows a consolidated figure and its
[stated scope](/learn/consolidated-group-scope) instead of a ratio.
## The evidence chain behind every figure
Every artifact behind the board — quorum logs, storage proofs, captured
PDFs, parsed reports, the methodology document itself, and the
pre-publication review — lands on an append-only, tamper-evident hash
chain, Merkle-anchored to external timestamping services nightly. The
methodology served today is the artifact with SHA-256
`08cbdb46beed306a430b35c4d6d5e1a03d2a1d1d29bfef65fa5105cf53820381`,
uploaded 2026-07-19; the board's figure routes withhold themselves if the
served document ever differs from the reviewed one.
## What to take away — and what we deliberately don't do
The takeaway is a habit, not a number: every figure should arrive with a
way to check it and a statement of what it does not say. That is the
whole methodology — proofs where the chain permits them, captured
documents before parsed ones, ratios only inside a single document's
stated scope, and an evidence trail that outlives any correction.
Just as deliberate is what stays out of scope. The board carries no
solvency verdicts, no price or depeg data, no predictions, and no
rankings that reach beyond what the documents themselves state. When a
figure is missing, that gap is published as the finding rather than
papered over — because the moment a measurement service starts filling
silences with estimates, readers lose the ability to tell which of its
numbers were ever measured at all.
## Disclaimer
ReserveBeat renders figures from issuer-published reports and public blockchains, verbatim and at their stated scope. It is not an auditor, makes no attestation, and states no opinion on any issuer's solvency. Nothing here is investment, trading, legal, tax, or accounting advice. No figure published here is a recommendation to hold, buy, sell, or redeem any token.
Tracked tokens are subjects of public data: supply figures are read from their public Ethereum contracts, and composition figures from the issuers' own published attestation reports. No tracked issuer is a customer and none has endorsed this site.
Source: https://reservebeat.com/research/how-reservebeat-measures-stablecoin-reserves