Two of the most quoted on-chain metrics measure exchange activity and routinely tell opposite stories in the same week. The reason is not that one is wrong. It is that one measures a flow and the other measures a level, and the things that break them are different.
This article is for informational and educational purposes only. It is not financial advice, investment advice, or a recommendation to buy, sell, or hold any security, cryptocurrency, or financial product. Always verify data with official sources before making financial decisions.
Short answer: a rate of change against a level
Short answer: Exchange netflow is coins moving onto exchanges minus coins moving off, measured over a period — a flow. Exchange reserve is the total balance held in identified exchange wallets at a point in time — a level. Netflow is noisy and responds quickly; reserve is smooth and responds slowly. Both depend entirely on wallet labelling that no analytics provider can verify with certainty.
How netflow is computed, and what breaks it
Providers maintain lists of addresses attributed to exchanges, sum inflows and outflows across them over a window, and report the difference. Positive netflow means coins are arriving; the standard interpretation is potential selling pressure, since coins must sit on an exchange to be sold there.
Three things break that interpretation.
- Internal transfers. Exchanges move funds between hot and cold wallets constantly. If a provider has labelled one address and not another, routine internal housekeeping registers as a large inflow or outflow that never involved a customer.
- Derivatives collateral. Coins deposited to post margin on a futures position are not being sold. They read identically to a deposit made in order to sell.
- Custody and ETF flows. Institutional custodians and exchange-traded product wallets are labelled inconsistently across providers, and large movements between them can dominate a day’s netflow with no market intent behind them.
Reserve totals and the custody migration problem
Exchange reserve has fallen substantially over recent years, and the standard narrative reads that as accumulation — coins moving to self-custody by holders with no intention of selling.
Part of it is that. A meaningful part is structural migration that has nothing to do with conviction. Institutional buyers increasingly hold through regulated custodians and exchange-traded products rather than on trading venues. Coins moving from an exchange wallet to a custodian wallet reduce measured exchange reserve while representing an increase in institutional holdings, not a withdrawal from the market.
There have also been genuine one-off events — exchange failures, regulatory actions, market structure changes — that moved large balances for reasons unconnected to any holder’s price view. A reserve chart smoothed over years hides all of them.
Reading netflow spikes alongside funding and open interest
Netflow in isolation is close to uninterpretable. Combined with derivatives data it becomes more useful.
A large inflow alongside rising open interest and elevated funding rates suggests coins arriving as collateral for leveraged positions rather than for sale — the position is being built, not exited. The same inflow with flat open interest and neutral funding is more consistent with spot selling intent.
This cross-check does not resolve the ambiguity, but it narrows it. Any confident reading of netflow that ignores derivatives positioning is ignoring the most likely alternative explanation.
Why providers disagree
Wallet attribution is the entire foundation of both metrics, and it is inferential. Exchanges do not publish their address lists. Providers build them through clustering heuristics, disclosed addresses, proof-of-reserve attestations and manual research.
Different methodologies produce different address sets, which produce different netflow and reserve figures for the same chain on the same day. Discrepancies between major providers are routine and occasionally large. When two analysts cite conflicting on-chain evidence, they are frequently both quoting accurately from providers that disagree.
The practical rule: pick one provider and track its series consistently. Comparing a netflow figure from one source to a reserve figure from another produces conclusions that are artefacts of methodology.
Risks, uncertainty, and limits
Neither metric measures intent. A coin arriving at an exchange may be sold, may be posted as collateral, may be moved for custody reasons, or may sit indefinitely. The metric records the transfer and nothing else.
On-chain data is transparent about transactions and opaque about purpose. Analysis built on it should carry that uncertainty explicitly rather than presenting attribution as observation.