Bitcoin flow analysis to trace fund movement on-chain

Each transaction input spends a specific earlier output, so funds can be followed step by step across hops and address clusters. BitcoinDatabase does this over the fully-indexed chain and summarizes exposure to labeled entities as context.

No. It surfaces on-chain flow data and risk signals to support a compliance team's own review. It makes no accusations, no identity proofs and no deanonymization claims, and it is informational data, not legal or financial advice.

It is following value across transaction hops to answer where funds came from or where they went. Each Bitcoin transaction spends specific previous outputs, so the chain records an exact edge list, and flow analysis walks it. The chain part is precise. What is inferred, and what people usually mean when they say the analysis is uncertain, is which real world entity sits behind each address.

To the coinbase transaction that created the coins, in principle, because every output traces to a previous output or to newly mined supply. The practical limit is not depth but branching. Hops multiply fast, and once funds pass through an exchange or a mixer the outgoing coins are no longer the incoming ones in any meaningful sense, so tracing past that boundary tells you about a service rather than about a person.

It should not, and forgetting this is the single most common error in on-chain flow numbers. Most Bitcoin transactions send the bulk of their value straight back to the sender as change. We measured one block where raw output volume was 10,285.76 BTC but only 2,113.47 BTC, about 20.55 percent, actually left the sender, an overstatement of 4.87 times. Six days earlier the same method gave 17.03 times on a different block.

Partly, and by design. CoinJoin and PayJoin deliberately break the common input ownership assumption that most clustering relies on, so inputs in one transaction no longer imply a single owner. Mixers, exchange hot wallets and Lightning channels all break the chain of custody in different ways. Flow analysis stays useful, but treat it as evidence to weigh, not proof.

Flow analysis follows value between addresses; clustering groups addresses believed to share an owner. You generally need both, because tracing address by address produces noise once change addresses proliferate, and clustering is what collapses that back into entities. Clustering is heuristic, so its output carries uncertainty that then propagates into any flow number built on top of it.

Match the receiving addresses against a labeled address set, then check what the chain actually shows. The catch is that attribution is inference: a deposit address is only known to belong to an exchange because somebody observed it behaving that way. We track this constantly on published reserve addresses and the labels are frequently disputed, so state your confidence rather than presenting an attribution as fact.

It is used as evidence, but it is not self-proving. The chain shows that outputs moved; it does not show who controlled the keys, whether a transfer was a payment or an internal rebalance, or whether an address label is correct. Courts and regulators treat it as one input alongside subpoenaed exchange records. Anyone presenting a flow diagram as proof of intent is overstating what the data supports.