How this is measured
What the numbers are, how they are produced, and what they cannot tell you.
Data provenance
Figures come from chain state and event logs, ingested and modelled into an analytical warehouse. The site reads those measurements through chart-specific queries, with independent onchain checks rather than protocol self-reporting. Charts use warehouse results or dated captures of those results, refreshed together so every chart page reports the same observation date; research articles keep the date they were written against.
Chain coverage
Supply and holder balances are measured on Ethereum, BNB Chain, Base, Arbitrum, Arc, Avalanche, Ink, Monad, Optimism, MegaETH and Etherlink. Bridge escrow is measured on Ethereum, Base, Arbitrum, Avalanche and MegaETH, so supply there is reported net of tokens held in escrow; on BNB Chain, Arc, Ink, Monad, Optimism and Etherlink escrow is not tracked, and supply is reported gross. A naive sum of gross issuance would count bridged tokens twice, while the current total combines netted and gross chain-level figures because escrow coverage is incomplete.
Wrapper classification
A wrapper is classified by redemption structure and custody model, not by marketing category. Tokens with the same ticker on different chains are separate instruments unless a canonical bridge makes them fungible. WBTC, cbBTC, LBTC and tBTC are shown by name; every other tracked wrapper — BTCB, FBTC, SolvBTC, uniBTC, renBTC and the rest of the covered universe — is aggregated into one series rather than dropped.
Protocol classification
Holder addresses are grouped by curated contract labels — lending market, bridge escrow, DEX pool, custodian, vault, exchange — resolved from a source-priority ladder in which reviewed internal labels outrank external aggregators and predictions. An address with no protocol label falls back to a wallet classifier that separates externally owned accounts from contracts. Unlabelled contracts stay in an explicit 'other' bucket; they are never distributed across the known categories.
Holder interpretation
Holder charts classify balances by the address or contract holding the tokens, not by beneficial owner. One entity can control several addresses, while one custodial address can hold balances for many entities. Address-level concentration therefore cannot establish beneficial-owner concentration. Holder shares are computed on gross balances, which is why bridge escrow appears as its own bucket there; supply netting depends on which chains have escrow coverage.
Limitations
Lending collateral covers the venues the warehouse indexes today — Aave v3, Morpho Blue and Spark — so it is a floor on collateral, not the whole lending market. Label coverage is incomplete by construction: what is not labelled is reported as unlabelled rather than guessed. Onchain data cannot see offchain claims, rehypothecation, or the terms of a bilateral agreement; where a conclusion depends on something not visible onchain, it is stated as an assumption in the article rather than embedded in a chart.
Update cadence
Where live queries are supported, production charts use warehouse results cached for six hours. Other charts, and failed live queries, use dated captures. Charts inside research articles are drawn from a copy of the data taken for that article, dated with the article's data-through date, so they do not move when the chart pages are refreshed; an article is re-issued against newer data only deliberately, with its text and dates updated. Each chart carries the date its own source runs through — different sources advance on different clocks, and a shared 'last updated' stamp would hide that.