Balances give the level and this gives the rate at which that level is changing, which means it turns first. A stock of coin cannot move without the flow into it moving beforehand, so the earlier reading is a matter of construction rather than of luck.

It is smoothed for a reason worth understanding. A single day of exchange flow is dominated by whichever large transfer happened to land that day, and reading those days one at a time is reading noise with great attention.

What it actually measures

The sign is the reading. Positive means more arriving than leaving across the period, negative means the reverse, and the magnitude matters far less than which side of the line the figure has settled on.

Agreement with the balance trend is what confirms it. Flow and level pointing the same way across weeks is a genuine movement; a single week set against a months-long trend has usually turned out to be noise.

Internal housekeeping is invisible here. Venues move coin between their own wallets constantly for reasons of their own, and some of that reshuffling is indistinguishable from a genuine deposit or withdrawal.

Arrival creates the possibility of selling and nothing more

Inflows get read as impending supply, which is understandable: coin has to be on a venue before it can be sold there, so arrival is a necessary step.

It is not a sufficient one. Coin arrives for collateral, for lending, for market-making inventory and for reasons nobody outside the venue can see, and a large amount of it never trades at all. Necessary and sufficient get conflated here more than anywhere else on the dashboard.

Flow Matrix: a venue story and a market story look alike in aggregate

The interesting reading is whether the venues agree. The whole complex moving the same way points to something market-wide; a few gaining while others bleed points to something about the venues themselves.

An aggregate flow reading cannot tell you whether a large outflow was the market withdrawing coin or one exchange losing its users to another. Both produce the same line.

Splitting by venue separates them, and the separation changes what the reading means entirely. Coin leaving one exchange for another has not left the market at all, and treating a migration as an accumulation signal is a specific and avoidable error.

What it does not tell you

It cannot distinguish purpose at all. A deposit made ahead of a sale, one posted as collateral and one funding market-making inventory produce an identical reading, and there is no way of separating them from outside the venue itself.

The smoothing that makes it readable also delays it. A genuine turn appears here later than it actually happened, which is the price paid for not reacting to every large transfer that lands.

It covers tracked venues only. Movement to and from anywhere unobserved is silent here, so a large flow between an observed exchange and an unobserved one reads as though coin had left the market entirely.

How to read it

Net inflows. More is arriving than leaving over the period being read.

Flat flows. Deposits and withdrawals are cancelling each other out.

Net outflows. More is leaving than arriving over the period.

The Exchange Holdings dashboard carries Net Flow together with Balances, Flow Matrix and Coinbase Premium.

Common questions

Why is this smoothed?

Because a single day of exchange flow is dominated by whichever large transfer happened to land. Smoothing shows the direction without one movement setting the whole reading.

Do inflows lead to selling?

They make selling possible, which is a weaker claim than it sounds. Coin has to be on a venue before it can be sold there, and plenty arrives and never trades.

What confirms a genuine trend?

Flow and the balance trend agreeing across weeks rather than days. One week pointing against a months-long trend is usually nothing at all.

Why does this turn before the balance does?

Because a stock cannot change unless the flow into it changes first. The earlier reading here is arithmetic rather than an empirical claim about how markets behave.

Can venue housekeeping distort it?

It can. Exchanges move coin between their own wallets constantly, and some of that is indistinguishable here from a genuine deposit or withdrawal.

Why score each venue against itself?

Because exchanges differ in size by orders of magnitude. A raw comparison would surface the largest few and hide an unusual month at a smaller one entirely.

Why does agreement between venues matter?

Because it separates a market story from a venue story, and the two call for entirely different conclusions. The complex moving together says something about the market; a few moving against the rest says something only about those exchanges.