A transfer to an exchange is usually read as a sale. Most of the time that is wrong: the token arrives at the exchange address from another exchange or from a liquidity pool, and the sale — if there was one — happened at a different step entirely. So the flow on this page is not merely added up but split by origin. Without that split, the list would lie.
Eight exchange addresses found among RNBW holders. The bar under each one breaks the inflow down by where the sender got the token in the first place. The windows switch; inside each exchange the largest senders unfold.
These are addresses, not owners: thousands of people stand behind one exchange hot wallet, and one person can send from several addresses. Served as /api/exflow.json.
For every large sender we look one step back: where did they get the token themselves. That single step is enough to tell movement between venues apart from an actual sale.
The sender received the token from another exchange's hot wallet — or is such a wallet itself. This is shuffling between venues: liquidity, settlement between exchanges, a user withdrawing from one exchange to another. There is no pressure on the price here, yet in a raw list such a transfer would look like a large sale.
The token was taken straight from a liquidity pool — Aerodrome or Uniswap — and carried to an exchange. This is arbitrage on the price gap between the decentralised and the centralised venue. There is a sale here, but it happens in the pool, not on the exchange, and counting it twice is not allowed.
The source has a non-zero balance of its own — it is not a pass-through address. This is the closest thing to what people usually mean by "selling": somebody held the token and carried it to an exchange.
Sometimes the source is empty itself: the token passed straight through, and the step back explained nothing. We call that share unknown rather than spreading it across the other three. Separate from it stands the "not traced" share — small senders we never got to. These are different things, and in the first measurement we conflated them ourselves: Gate.io came out at 50.5% "unknown", when almost all of that share meant not "the chain runs further" but "we did not look".
Why one step is enough. The chain can be followed a long way, and each further step adds less: the token ultimately comes from a pool, from the mint, or from another exchange, and the answer to "where is the pressure from" already arrives at the first step. Each step, meanwhile, costs a full address history — time and quota. So the step is one, and the unexamined share is shown honestly instead of hiding behind a neat round breakdown.
A token carried onto an exchange may simply sit there for months. The reverse happens too: what gets sold is what has been sitting on the exchange since last year, and no inflow shows up that day at all. Inflow is an intention to trade, not a trade.
Every exchange has flow in both directions, and inflow without outflow is half the picture. This page deliberately counts only the way in: that is what answers the question asked — who sends. Net flow per exchange will become its own page when there is something to say beyond arithmetic.
An exchange is recognised by its address label, and labels are not known for every venue. If some exchange is labelled in no public source, its address is simply absent here — and we will not find out. The list of exchanges was checked by hand rather than assembled by keyword: among the labels containing "Coinbase" there are ordinary user wallets too.
Recomputed every six hours. Origins once established are remembered, so only the first pass is expensive.