Every so often someone shows me a netflow chart and asks why nothing is moving. Exchange balances flat, netflow hovering around zero, no obvious pressure either way, and yet the price has been grinding up for weeks. The assumption is always that the chart is broken or the data is lagging. It usually is not. What is happening is that the biggest buyers are settling somewhere the chart does not see, and by design. Once you understand how an over-the-counter desk actually fills a large order, the flat netflow stops being a mystery and starts being a signal in its own right.
Why big money avoids the order book
Say a fund wants to buy a few thousand coins of something with a thin book. If they walk that order into a public exchange, they eat the book. The first slice fills near the screen price, and every slice after that fills worse, because they are the ones lifting each level. By the time the order is done they have moved the price against themselves and told the entire market what they are doing. Anyone watching the tape front-runs the rest of the fill. This is slippage and information leakage, and at size both are brutal.
An OTC desk exists to make that problem go away. Instead of hitting the book, the buyer negotiates a single price for the whole block directly with the desk, off-exchange. The desk quotes a price, they agree, and the trade settles bilaterally. The buyer gets one clean fill with no visible footprint on any order book, and the desk takes the other side, either from its own inventory or by sourcing the coins elsewhere. The desk manages the risk of unwinding its position quietly over time, which is the service you are paying the spread for.
The important part for anyone reading flow data is that none of this touches the exchange's matching engine. There is no trade printed on the public tape, no candle that reflects the block, no dent in the visible book. From the outside the market looks calm while a large position quietly changes hands.
What settlement actually looks like on-chain
Here is the thing that trips people up. OTC does not mean invisible. It means invisible to exchange flow data, which is a much narrower claim. The coins still have to move, and on-chain that movement is right there if you know what to look for.
A typical desk settlement leaves a recognizable shape. You will often see a large transfer from a known desk wallet or a custodian into a fresh address, or a series of round-number transfers that do not correspond to any exchange deposit or withdrawal. The counterparty side frequently moves into cold storage or a qualified custodian rather than back onto an exchange, because a fund that just accumulated a block is not planning to sell it next week. So the flow you would normally read as neutral is actually two large wallets doing a handoff that never went near a public book.
The tells I pay attention to:
- Repeated round-number transfers between the same cluster of addresses that never deposit to or withdraw from an exchange.
- Large outflows from a desk-associated wallet landing directly in a custody address rather than a hot wallet.
- Netflow reading flat or slightly positive while stablecoin balances on the buy side quietly drop, which is the cash leg of the settlement.
- Accumulation addresses that only ever receive and rarely send, growing steadily while exchange balances do nothing.
None of these are proof of an OTC block on their own. Labeling is imperfect, desk wallets get rotated, and custodians commingle client funds so a single address can represent many parties. But the pattern of large, book-avoiding transfers into sticky custody is a very different picture than the flat netflow chart suggests, and it is usually the more honest one.
Why flow-based conclusions break during accumulation
The failure mode I see most is someone reading flat or neutral exchange netflow and concluding there is no demand. During a genuine institutional accumulation phase this reasoning is backwards. The absence of visible exchange buying is not the absence of buying. It is what serious buying looks like when the buyers are competent enough to use a desk. If anything, a long stretch of suspiciously calm netflow alongside a grinding price is one of the more reliable hints that the accumulation is happening off-book.
Exchange netflow is genuinely useful for a lot of things. It is good at catching retail-driven inflows before a sell-off, at showing you when coins are leaving exchanges into self-custody in aggregate, at spotting panic. What it is bad at is measuring the largest and most deliberate flows in the market, precisely because those flows were structured to avoid it. Treating a metric as complete when its blind spot is exactly the participant you care about is how you end up confidently wrong.
A workflow that does not get fooled
I do not throw out netflow. I just refuse to read it alone, especially when the price action and the flow disagree. When those two things disagree, the flow is usually the one missing context.
The rough process I run:
- Start with exchange netflow as the baseline, but treat a flat reading during a trending market as a question, not an answer.
- Pull up the wallets you can label. Known OTC desk addresses, large custodians, and repeat accumulation clusters. Watch for handoffs between them that skip exchanges entirely.
- Check the custody direction. Coins moving into cold storage and staying put is a very different story than coins cycling through hot wallets.
- Cross-check the stablecoin side. A block trade has a cash leg, and stablecoin movement that lines up in size and timing with a token transfer is a good corroborating signal.
- Only then form a view, and hold it loosely, because address labeling is the weakest link and you will be wrong about specific wallets more often than you would like.
This is roughly the discipline we try to bake into how we present flow at Blockcircle, pairing exchange-level data with wallet-level tracking so the desk-shaped hole in netflow is at least visible rather than silently missing. The point is not to catch every OTC block. Nobody does. The point is to stop trusting a single metric during exactly the phase when that metric was engineered to lie to you. When netflow reads neutral and the price keeps climbing, do not assume nothing is happening. Assume the interesting part moved somewhere the order book cannot see, and go look for the handoff.