On-chain metrics

Definition

On-chain metrics are indicators computed from the public ledger itself — a data surface no other asset class has: every transfer, address balance, and fee is permanently observable (NIST IR 8202). Standard families: activity (active addresses, transfer counts/volume), holder behavior (dormancy/coin-age, long-term-holder supply), flow proxies (exchange inflows and outflows), and valuation ratios built on them (market cap over on-chain volume and similar). The promise is fundamental-style analysis for an asset with no cash flows; the documented caveat is that entity ambiguity, internal churn, and concentration make raw metrics far noisier than their dashboards suggest (Makarov-Schoar 2021).

How it works / structure

• What the ledger gives: address-level balances and transfers, timestamps, and fees — public, complete, and replayable (crypto-transfer-settlement). What it does NOT give: identity. Addresses are not people; one entity spans many addresses and one custodial address spans many people.
• Entity clustering: analytics vendors merge addresses into presumed entities by heuristics; Makarov-Schoar (2021), clustering the bitcoin ledger, found holdings and mining highly concentrated and much measured volume attributable to exchange-internal and inter-exchange churn rather than economic transfers — the core measurement caveat.
• Exchange flow metrics: inflow to identified exchange addresses reads as sell-availability, outflow as withdrawal to custody (crypto-custody-models) — contingent on the vendor’s address labels being current and complete (dated, proprietary inputs).
• Holder-age metrics: coin dormancy and “supply last moved N years ago” separate long-holders from active supply (crypto-wallets-keys — dormancy conflates patience with lost keys).
• Custody-era decay: ETP-era institutional flows move inside custodians and OTC desks with thin on-chain footprints (crypto-etps) — the metrics’ coverage of economically meaningful flow is shrinking, not growing.

When it applies

Crypto-native supplements to sentiment and positioning analysis (the closest analogs to sent-fund-flows / sent-short-interest for coins), usage-vs-price divergence theses (crypto-coins-vs-tokens), and supply-context work (crypto-supply-schedules). Every use inherits the platform’s data-hygiene bar (quant-data-hygiene): vendor, labeling methodology, and revision policy pinned before the metric enters evidence.

Risk profile & failure modes

• Entity illusion: treating addresses as investors — the first-order error; concentration and churn findings (Makarov-Schoar 2021) invalidate naive activity readings.
• Label drift: exchange-flow metrics silently degrade as venue wallet architectures change; a metric’s history is only as stable as the vendor’s relabeling policy (revisions rewrite the past — replay hazard).
• Goodhart pressure: publicized on-chain signals invite manufactured activity (self-transfers are nearly free) — visibility cuts both ways.
• Backtest overfit: many published on-chain “valuation bands” are in-sample narratives without out-of-sample evidence — folklore-adjacent until replayed (quant-backtest-hygiene).

Evidence & limits

The ledger’s observability is protocol fact; concentration and churn measurements are NBER-published (Makarov-Schoar 2021). Predictive claims for specific on-chain metrics lack robust peer-reviewed support — the platform treats each as a candidate signal requiring its own replay, with vendor-methodology pinning, and labels vendor loss/flow tallies as industry data.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Sustained net exchange outflows (30-day) precede positive 90-day returns at better than coin-flip rates over the replay sample (flow-signal thesis)” — falsified by the replay.
• “Active-address growth and price diverge (rising price, flat-or-falling activity) for two consecutive quarters (usage-price divergence thesis)” — falsified by the paired series.

Cross-references

• What generates the data: crypto-transfer-settlement; what it measures: crypto-coins-vs-tokens, crypto-supply-schedules
• Sentiment/positioning analogs: sent-fund-flows, sent-short-interest, sent-news-social
• The discipline that gates it: quant-data-hygiene, quant-backtest-hygiene