On-Chain Analytics

Active Addresses and Entity Adjustment

Active addresses count the addresses that appear in a period's transactions. The raw count is a fact about the ledger, but it is a poor measure of how many people are using Bitcoin, because one person can control thousands of addresses. Entity adjustment replaces addresses with clusters, and that replacement is an inference.

Last reviewed 2026-09-21Source: Coin Metrics address and entity-adjustment methodology; Bitcoin Core UTXO setNo figure is quoted here. The raw count is a ledger fact; the entity-adjusted count depends on address clustering.

Written by: The Bitcoin Data Guide editorial team
Reviewed by: Data review desk
Dates: Published 2026-09-21 · Last reviewed 2026-09-21

Fact

The ledger records every address that appears in a transaction, so the raw active-address count is directly computable.

Forecast

Reading a rising active-address count as growing adoption, or an entity-adjusted count as the number of users, is an interpretation that the clustering heuristics do not establish.

The raw count

An active address is one that appears as an input or an output in at least one transaction during a period. Counting the distinct addresses that appear gives the active-address count for that period. The count is a fact about the ledger: the addresses are recorded, and the count is arithmetic on top of them.

The count is offered as a measure of network activity, and it is often read as a proxy for the number of users. That reading is where the measure breaks down. One person can control an unlimited number of addresses, and a single wallet may generate a new address for every payment it receives. A service that pays out to many addresses can inflate the count without any new user arriving.

The count is also sensitive to how addresses are reused. A wallet that reuses one address produces a low count; a wallet that generates a fresh address per payment produces a high one. The same economic activity can therefore produce very different counts depending on wallet software, which is a property of the software rather than of the network's use.

What is observed and what is inferred

What the ledger directly records is which addresses appear in which transactions. The raw active-address count is therefore a fact about the chain, and so is the transaction count and the transfer volume. None of them requires an assumption beyond the ledger's own structure.

The inference begins when the count is read as a number of users. The step from "this many addresses were active" to "this many people were active" requires knowing which addresses belong to the same owner, and the ledger does not record ownership. Entity adjustment is the attempt to make that step: it groups addresses into clusters and counts clusters instead of addresses.

The entity-adjusted count is therefore an inference built on the clustering. It is a better proxy for user activity than the raw count when the clustering is right, and it is a worse one when the clustering is wrong, because it hides the raw data behind a model.

The assumptions it depends on

Entity adjustment depends on address clustering: the heuristics that decide which addresses belong to the same owner. The common ones rest on assumptions about how wallets construct transactions — that inputs spent together share an owner, that change returns to the sender, and that a fresh address belongs to the wallet that funded it. Each assumption is a reasonable default and each can be defeated by a wallet designed to defeat it.

Entity labelling is a further inference on top of the clustering. It attaches a name to a cluster, and a mislabel propagates into every metric built on it. A cluster labelled as an exchange may belong to a payment processor or a custodial service, and the error is invisible in the adjusted count.

The most common distortion is the growth of batching and internal settlement. A service that batches many payments into one transaction reduces the transaction count without reducing the activity, and a service that settles internally never touches the chain at all. The adjusted count sees only the on-chain portion of activity, which is a shrinking share of the total. The tracing page sets out the clustering heuristics in full, and the limits of blockchain analysis page explains why a cluster is an inference rather than a fact.

Sources and references

The description of the measures and their construction is taken from the published methodology of the providers that compute them, and from the academic work on address clustering.

The activity measures and their methodology. Coin Metrics, Active Addresses: documents the count and the entity-adjustment procedure.
The clustering heuristics behind entity adjustment. Meiklejohn et al., A Fistful of Bitcoins: characterises the address clustering that entity adjustment depends on.
The change-output heuristic and its limits. Bitcoin, Bitcoin Developer Guide — Transactions: describes how change outputs are created, which is the basis for one of the most common clustering assumptions.

Sources

Coin Metrics address and entity-adjustment methodology: Documents the active-address count and the entity-adjustment procedure built on address clustering.
Bitcoin Developer Guide — Transactions: Describes how addresses appear in transactions, which is what the count reads.
A Fistful of Bitcoins: Characterises the address clustering and attribution assumptions that entity-adjusted metrics depend on.