Exchange inflows and outflows: how to read coins moving to and from exchanges

What counts as an exchange inflow or outflow?#

Centralised exchanges hold customers’ coins in their own wallets: deposit addresses, hot wallets for day-to-day withdrawals and cold wallets for storage. When you deposit bitcoin, you send it to an address the exchange controls. That on-chain transfer is visible to everyone — but nothing on the blockchain says “this address belongs to Exchange X.”

Data providers therefore build lists of exchange addresses. Coin Metrics counts coins as sent to an exchange when they go to an address it identifies as exchange-owned, and as withdrawn when they leave the control of such an address ^[1]. Glassnode describes three ways of finding those addresses: addresses verified by exchanges (for example in proof-of-reserves disclosures), public address tags checked for quality, and clustering algorithms that keep extending the label set ^[2].

How is exchange netflow calculated?#

For Bitcoin there is a subtlety. A single transaction can both withdraw from and deposit to the same exchange — for instance when change goes back to an exchange wallet. Coin Metrics nets these inside the transaction: if a transaction sends 90 BTC to exchange A and also withdraws 50 BTC from it, the flow is recorded as +40 BTC, not as +90 and −50 ^[1]. Change outputs exist because Bitcoin outputs must be spent in full ^[3]; see the netflow metric profile for more detail.

Daily netflow swung between −1,000 and +1,300 BTC, yet the week netted to just +50 BTC. Single days are noisy; a running total says more about the direction of travel.

Running total of the hypothetical week’s netflow

Cumulative netflow (BTC)

How do analysts usually read inflows and outflows?#

There is no rule linking flows to price, only common interpretations. The most widespread is a hypothesis about intent: coins usually need to be on an exchange to be sold there, so sustained net inflows are often read as supply being made available for sale, and sustained net outflows as coins moving into self-custody. Treat that as a starting hypothesis to test, not a fact the data proves.

Glassnode itself warns that early signs such as a significant inflow or outflow should be approached with caution, because exchanges can change addresses and reshuffle funds internally, and some of these events cannot be detected straight away ^[2].

Why are exchange balances only estimates?#

Everything depends on the address list. Coin Metrics says its exchange supply figures should be seen as an underestimate, because its heuristics and sources might not discover every address an exchange owns ^[4]. Glassnode says its balances can largely be considered lower bounds, since it tries to avoid falsely labelling non-exchange addresses ^[2].

• Coverage gaps. Providers only track the exchanges and chains they support; Coin Metrics withholds a metric when it judges its coverage of an exchange incomplete ^[1].
• Retroactive revisions. When a new exchange address is found, past balances can be recalculated. Glassnode recommends its point-in-time data for anyone who needs values that never change ^[2].
• Disclosure differs by exchange. Accuracy is better where an exchange publishes verified addresses; Glassnode’s own table shows that many exchanges do not ^[2].
• Assets only, not liabilities. On-chain balances say nothing about what an exchange owes its customers ^[2].

How should a beginner use exchange flow charts?#

1. Check the coverage

   Which exchanges and chains are included? An “all exchanges” line only covers the exchanges that provider tracks.

2. Look at totals over weeks

   Use a running total or a 7-day sum rather than a single day’s netflow.

3. Compare with exchange balance

   Netflow and the change in exchange balance should tell the same story. If they do not, a relabelling may be involved.

4. Ask what else could explain it

   List the non-selling reasons for the flow before accepting the selling story.

5. Keep it descriptive

   Say “coins left labelled exchange wallets”, not “investors are bullish”. The second is an opinion.

What mistakes do beginners make here?#

• Treating one day’s inflow as a sell signal

  Single-day spikes are often internal reshuffles or relabellings. Look at multi-week totals.

• Assuming the provider sees every exchange

  Coverage is limited to tracked exchanges, and exchange balances are underestimates by design.

• Confusing on-chain flows with trading volume

  Trades inside an exchange never touch the blockchain. Flows only show deposits and withdrawals.

• Mixing providers in one comparison

  Each provider has its own address list. Compare a series with its own history, not with another provider’s numbers.

Frequently asked questions#

The bottom line#

Exchange flows answer a narrow question — how many coins arrived at and left wallets believed to belong to exchanges — and they answer it with estimated labels. Used over weeks rather than days, and alongside the reasons coins might move for non-trading purposes, they can describe how supply is shifting. They do not reveal intentions.

For the exact definitions, read the exchange netflow profile, and for the bigger picture of what labels can and cannot do, see the limits of on-chain data.

Sources#