Public chains produce measurable data, and the standard metrics derived from it require care about what is actually being counted.
Active addresses
Counts addresses transacting in a period.
Which is not a user count, since one person can control many addresses and one address can serve many people.
Exchange addresses in particular represent enormous numbers of users behind a single entry.
Transaction count
Counts transactions rather than economic activity.
Which is inflated by automated activity, by protocol operations and, on low-fee networks, by activity generated for its own sake.
Transaction value transferred is a different and generally more informative figure.
Total value locked
The value of assets deposited in protocols.
Which double counts where assets are deposited, wrapped and deposited again.
It also rises and falls with asset prices independently of any change in deposits, which is frequently misread as growth.
Exchange balances
Tracked as an indicator of assets available for sale.
Which depends on correct identification of exchange addresses.
Reattribution of addresses causes apparent large movements that reflect labelling changes rather than transfers.
Realised value
Valuing each unit at the price when it last moved rather than at the current price.
Which produces a measure of aggregate cost basis.
Derived ratios comparing it to market value are widely used and depend entirely on the assumption that movement equals a change of ownership.
Dormancy and age
How long units have been unmoved.
Which is genuinely observable and is interpreted as holder behaviour.
Custodial reorganisation moves large quantities without any change in beneficial ownership, which distorts these measures.
Hash rate
Estimated from block production rate rather than measured directly.
Which means short-term figures are noisy and should be read as averages.
Apparent sharp changes over hours are generally statistical variation.
Using metrics honestly
Ask what is actually counted, what is double counted, and what changes when nothing real changes.
Most misuse of these figures comes from treating a proxy as the thing it proxies for.
This is methodological description and is not investment advice.
Attribution methodology
Analytics providers use heuristics to label addresses, and these differ between providers.
Which is why the same metric differs across sources.
Methodology documentation exists for the better providers and is worth reading before citing a figure.
Survivorship in protocol metrics
Deposits measured across protocols exclude those that failed.
Which produces an upward bias in historical series.
The same problem affects yield statistics quoted over long periods.
Cross-chain double counting
An asset bridged and deposited appears in figures on both chains.
Which inflates aggregate totals substantially.
Whether a series adjusts for this is a methodological choice that changes the number materially.
Correlation and causation
Metrics correlating with price historically are widely presented as predictive.
Which is a reasonable hypothesis and is not a demonstration.
Out-of-sample performance is what would test them, and it is rarely presented.
Using data well
Prefer metrics with clear definitions, check the methodology, and treat derived indicators with more caution than raw counts.
Fee revenue
What users actually paid to use a network or protocol.
Which is among the more meaningful figures, since it reflects willingness to pay.
Distinguishing fees paid to the protocol from fees paid to liquidity providers matters for interpretation.
Supply distribution
Holdings by address size band, published for most major assets.
Which shows concentration and is distorted by exchange and contract addresses.
Adjusted versions excluding known custodial addresses are more informative and depend on labelling quality.
Stablecoin supply
Aggregate issued supply as an indicator of capital present in the system.
Which is directly observable and requires care about which chains and which issuers are counted.
Issuance and redemption are visible on chain, which makes this among the cleaner series available.
Developer activity
Repository commits and contributor counts as a proxy for ongoing work.
Which is easily gamed and is still weakly informative in aggregate.
Commit counts say nothing about the substance of what was committed.
Using these responsibly
State what a metric measures, acknowledge what it does not, and avoid presenting proxies as direct observations.
Where to find the data
Public query platforms allow anyone to compute these figures directly from chain data.
Which means claims can be checked rather than accepted.
Community-published queries are generally visible, so the methodology can be inspected.
Closing
The transparency is genuine, and it only produces understanding when the definitions are read carefully.
Presentation and incentive
Metrics are frequently presented by parties with a position in the asset being described.
Which is not disqualifying and is relevant context.
Checking a striking figure independently takes minutes with public query tools.
The general rule
A metric is only as good as the definition behind it, and the definition is usually available if you look.
Treating any of these figures as a direct measurement of human behaviour rather than as a proxy for it is where most misinterpretation begins, and the correction is simply to name what is counted.