An on-chain chart is not the blockchain itself. It is a model built from address sets, entity labels, time boundaries, price assumptions and transformations. Use this framework before citing any on-chain metric in research.
The minimum definition record
| Field | What to record | Why it matters |
|---|---|---|
| Metric name | Exact vendor name and version | Similar labels can use different formulas |
| Unit | Coins, USD, addresses, entities, transactions or ratio | Prevents category errors |
| Address set | Included chains, contracts, wallets and exclusions | Coverage changes the result |
| Entity model | How addresses are clustered or labelled | Entity metrics depend on inference |
| Time boundary | UTC, exchange day, rolling window or block interval | Daily values can differ by cutoff |
| Price source | Venue, index, close, VWAP or provider | USD-denominated metrics inherit price methodology |
| Transformation | Smoothing, adjustment, realized value or cohort rule | The model may be farther from raw events than it appears |
| Revision policy | Whether history changes after relabelling | Published figures may not remain identical |
Five questions before interpretation
- What is directly observed? Identify the raw transaction, balance or block-level event.
- What is inferred? Separate labels, ownership assumptions and behavioural cohorts.
- What is transformed? Record smoothing, aggregation, price conversion and normalization.
- What is excluded? Identify unsupported chains, venues, wrapped assets or internal transfers.
- What would falsify the interpretation? State the alternative mechanism that could move the same chart.
Common metric traps
| Metric | Tempting conclusion | Required caution |
|---|---|---|
| Exchange inflows | Selling is imminent | Transfers do not reveal owner intent |
| Stablecoin supply | Buying pressure is coming | Supply is capacity, not intention |
| Active addresses | Users are growing | One user can control many addresses |
| Realized price | Investor cost basis | Last coin movement is not necessarily purchase price |
| Smart Money | Expert conviction | The cohort depends on labelling and selection rules |
| Protocol revenue | Token value accrual | Protocol income and tokenholder rights are different |
Cross-provider comparison
When two providers disagree, do not average the numbers automatically. Compare venue and chain coverage, address clustering, asset mappings, time boundaries, price sources and historical revisions. Both providers may be internally correct under different definitions.
Publication record
| Field | Research note |
|---|---|
| Vendor and metric | |
| Definition link | |
| Query time | |
| Observation window | |
| Raw source or cross-check | |
| Known limitations | |
| Interpretation and alternative explanation |
Minimum citation standard
A published claim should preserve the provider, metric definition, observation period, query time and any relevant revision note. A screenshot without those fields is weak evidence.
Continue with the On-Chain Analytics & Market Data Guide and the analysis of why strong Ethereum on-chain data is not automatically a buy signal.