How to read any figure here: value, period, source, time
Every number on this site carries its period, its feed and its observation time, because a value on its own cannot be checked or compared.
Every figure on this site is displayed with four things attached: the value, the period it covers, the feed it came from and the moment it was observed. A bare number cannot be checked, reproduced or lined up against another number, because two figures that look identical can be measurements of different things taken at different times from different feeds. The four labels exist so that a reader can tell the difference without having to trust the site.
The value, and the units it is expressed in
Most figures here appear in United States dollars, and a dollar figure for a digital asset always contains a second measurement hidden inside it: a conversion rate. A 30-day fee total such as fee revenue over 30 days is the sum of each day's fees in the network's own token, each day converted at that day's reference price. It is not the 30-day token total converted at today's price, and the two can differ substantially when the token price has moved. The convention is stated because the choice changes the number.
Some figures are shown in native units instead, because converting them would destroy information. Staked supply and circulating supply are counts of tokens; hash rate is expressed in exahashes per second; peg deviation is in basis points because the quantity of interest for a stablecoin is a small distance from a target rather than a level. Where a figure is missing, the cell is empty rather than zero. An empty cell means the feed does not cover that asset or the metric is undefined for it; a zero means the measured quantity was zero. Treating one as the other is the single most common way to produce a wrong ranking.
The period: levels, flows and annualized figures are three different kinds of number
A level is measured at an instant. Market capitalization, total value locked, total supply and validator count are levels: they describe a state, and they have a timestamp but no window. A flow is measured across a window and is meaningless without one. Fees over 24 hours, decentralized exchange volume over 24 hours and 30-day volume are flows. Dividing a level by a flow, which is what most valuation ratios do, produces a number whose unit is time, and that is worth remembering when the ratio seems to have an intuitive meaning that it does not have.
Annualized figures are a third category and the most frequently misread. Annualized fees takes a trailing window and scales it to a year. Nominal staking yield expresses a current issuance rate as a per-year rate. Neither is a forecast, and neither becomes more reliable by being expressed per year — scaling a volatile 30-day window by twelve multiplies the noise along with the signal. Change metrics such as the 30-day change in total value locked and 30-day fee growth are differences between two observations, so both endpoints carry their own timestamps, and a change figure computed across a gap in coverage is suppressed rather than estimated.
The source determines how far a number can be verified
Three kinds of feed produce the figures here, and they do not deserve equal confidence. The first is on-chain data, read from a node or an indexer. Transaction counts, fees paid, issuance, burns, validator sets and contract balances belong here. Anyone running the same software over the same blocks reaches the same answer, so a disagreement is a bug that can be found.
The second is off-chain data reported by trading venues: prices, volumes and order activity. A centralized exchange keeps its books internally and publishes what it chooses. Reported volume in particular has a documented history of inflation through wash trading, and no external party can fully audit it, which is why volume-derived figures such as turnover carry the same caveat as the volume they are built from.
The third is derived or modeled data, produced by this site or by an external data provider from inputs of the first two kinds. Real staking yield, which subtracts supply growth from the nominal rate, is derived. So is every ratio on the valuation tab. Derived figures are reproducible only if the formula is published, which is why each one links to its entry in the metric catalog and from there to the methodology page.
The observation time is not the same as the period
The period says what stretch of the world a figure covers. The observation time says when the figure was read. A 24-hour fee total observed six hours ago covers a window that ended six hours ago, and a reader who mixes it with a level observed one minute ago has built a comparison across a seven-hour gap without noticing.
Chain figures are anchored to block height rather than to wall-clock time, because block height is the only unambiguous position on a chain. The site indexes to a height a fixed distance behind the tip, because recent blocks can be replaced. A chain reorganization discards blocks that were briefly canonical, and networks differ in how quickly finality is reached — some declare a block irreversible within an epoch, others make reversal progressively more expensive without ever declaring it impossible. Reading at the tip would produce figures that quietly change afterward. Venue figures are anchored to trade time in coordinated universal time, and disclosure-based figures such as coins held by exchange-traded products are anchored to the publication date of the disclosure, which is typically at least one business day behind the position it describes.
Revisions, and why a chart can change shape after publication
Published figures are sometimes restated. An indexer corrects a decoding error and a month of fee history moves; a venue republishes a corrected volume series; a company files an amendment. When that happens the site records a revision rather than overwriting history silently, and a backfill that extends coverage into the past is marked as such so it is not mistaken for newly generated activity.
This matters beyond bookkeeping. Any study of how a metric behaved before an event has to use the values that existed at the time, not the values as later corrected; using corrected values introduces look-ahead bias, and the resulting series describes a past that nobody could have observed. Keeping the observation time on every figure is what makes it possible to tell a revision apart from a change in the world.
The clearest way to see the four labels working together is to open any figure's entry in the metric catalog, where the definition, formula, feed and cadence are stated in one place, and then follow the feed link into the data sources page. The screener article that follows this one puts the same labels to work on a filter across every covered asset.