Order Flow Trading Explained

What is order flow trading?

Order flow trading is the practice of reading the executed transactions and the resting limit orders behind a price move, instead of inferring intent from the shape a move leaves on a candlestick chart. Every print an exchange publishes carries a price, a size, a timestamp and a flag that reveals which side crossed the spread. Order flow analysis turns that stream into a statement about who was in a hurry and who was being paid to wait.

A candle is a compression of that stream. Open, high, low, close and total volume are four price samples and one scalar; the order in which the volume arrived, which side initiated it, and whether it was met or absorbed are all discarded. Two bars with an identical body and identical volume can be built from opposite internals — one where buyers lifted offers all the way up and the offer kept refilling, and one where a single large passive bid soaked up sellers until they ran out of size.

The practical claim of order flow is narrow and worth stating plainly: it does not tell you where price is going. It tells you what just happened at a level with far more resolution than price alone, which is why it is normally used to confirm or invalidate a structural read rather than to originate one. Without a structural frame — where the range is, which side of it price is on, what liquidity sits above and below — order flow gives you a very high-resolution view of nothing in particular. Market structure comes first; order flow decides whether the structure is being defended.

Where does order flow data actually come from?

From two feeds, both published by the exchange itself: the trade tape and the order book. The tape is the record of executions, one message per trade or per aggregated trade, carrying price, size, timestamp and the maker/taker flag. The book is the set of resting limit orders at each price, delivered as a snapshot plus a stream of incremental updates.

Every order flow layer is a derivative of those two feeds. Delta and cumulative volume delta come from the tape alone. Footprint charts come from the tape bucketed by price and by bar. Heatmaps, book walls and depth-of-market ladders come from the book. Absorption is one of the few readings that genuinely wants both at once: aggressive volume from the tape, replenishment from the book.

One detail that catches people out: many exchanges publish aggregated trades, collapsing all the fills of a single taker order against multiple makers at the same price into one print. That is convenient for bandwidth and misleading for size statistics. A large-print detector is measuring aggregated prints, not orders, and on venues with different aggregation rules the same underlying activity produces different print-size distributions. Any whale or large-trade layer has to say which of the two it is counting, or its threshold means nothing across venues.

Aggressor side: how the taker is identified

Exchanges do not publish "this was a buy". They publish which side was the maker, and the aggressor is the other one. On Binance, each aggregate-trade message carries a boolean field m meaning "the buyer was the market maker": when it is true, the resting order was a bid and a seller crossed into it, so the aggressor was the seller; when it is false, a buyer lifted a resting ask.

That single flag is the entire basis of signed volume. Signed volume summed per bar is delta. Delta accumulated across bars is CVD. Delta split by price level inside a bar is a footprint cell. Invert the flag and every layer downstream inverts with it, silently and plausibly — the chart still looks like a chart. It is also why a system consuming both a spot feed and a perpetual feed must apply the identical convention to both; otherwise the spot-versus-perp comparison, which is one of the more useful reads in crypto, degrades into noise.

Where no maker flag is published, the aggressor side has to be estimated. The classic approach compares the trade price against the prevailing quote: at or above the ask, classify as buyer-initiated; at or below the bid, seller-initiated; inside the spread, fall back to the direction of the last price change. This works acceptably on liquid instruments and poorly in fast markets, where the quote you are comparing against is already stale. Hold onto the distinction: a published maker flag is a measurement, a quote rule is a model, and models have error bars that the resulting delta chart does not display.

The three views of order flow

Order flow is looked at through three surfaces, and each one answers a different question. They are complements, not alternatives — the common beginner mistake is to adopt one and expect it to carry the whole read.

A book without a tape tells you what is being offered but not what is being taken. A tape without a book tells you what was taken but not what was left standing. The footprint is the join of the two along the price axis: it is the tape, sorted by price, arranged inside the bar you already look at. That is why footprint tends to be where discretionary readers start — it puts order flow on the same grid as the chart, so the question stops being "what does the tape say" and becomes "what happened at that specific level".

What does order flow show that a candle hides?

It shows initiative and distribution: which side spent aggression, at what price inside the bar, and what they got for it. A candle reports the net displacement; order flow reports the effort that produced it. The gap between the two is where most of the information lives.

Take a bar that travels up, prints a large positive delta, and closes back near its open. In candle form that is an indecisive doji with high volume — ambiguous. In order flow terms it is unambiguous: buyers crossed the spread in size, the offer met all of it, and price ended where it started. Effort without result. The mirror case — heavy negative delta into a low with a close in the upper half — is the standard shape of sellers being absorbed.

The same logic explains why volume alone was never enough. Volume is unsigned; it tells you a transaction happened, not who needed it. Two participants are present in every trade, and the interesting question is always which of them chose the price and which of them chose to wait. Order flow answers exactly that question and nothing beyond it, which is both its power and its ceiling.

Why do spot and perpetual order flow disagree?

Because they are different instruments with different populations. A perpetual swap is leveraged, funded, and liquidatable; its tape carries directional leverage, hedging flow and forced closes. A spot tape carries balance-sheet transfers: someone actually takes delivery of the asset. The two can pull in opposite directions for long stretches without either being wrong.

That divergence is a read in itself. When perpetual aggression is persistently buying while spot aggression is persistently selling, leveraged participants are lifting offers into holders who are handing them inventory. The mechanics of the funding rate and of open interest sit on top of this — see open interest for how positioning is measured alongside it — and the combination is more informative than any of the three alone.

This is also a reason to be careful with single-venue conclusions. A crypto asset trades on many exchanges plus perpetuals on more; one venue's tape is a sample, not the market. The sample is usually representative when the venue holds the bulk of the volume, and stops being representative exactly when something unusual is happening somewhere else.

What order flow cannot tell you

It cannot see what was never displayed and never executed, it cannot tell you why someone crossed the spread, and it cannot turn a description of the past into a forecast. These are structural limits, not tooling gaps, and any presentation of order flow that does not state them is selling something.

Specifically:

• Hidden and iceberg liquidity is invisible by construction. You infer it after the fact, from the mismatch between what displayed and what executed — never directly.
• Aggressor is not intent. A market sell can be a liquidation, a delta hedge, a rebalance or a mistake. The tape records urgency, not motive.
• One venue is a sample. Cross-venue and spot-versus-perp comparison exists precisely because a single tape can mislead.
• It is descriptive. Order flow tells you the state of an auction now. Whatever you do with that is a separate discipline, and it is not one this site engages in: we publish analysis, never instructions.

How order flow is read on The Confluence Show

The Confluence Show is a live, AI-run market analysis broadcast. Its analyst, NAIRO, reads raw trades, order books and positioning 24 hours a day through the Confluence Engine — more than forty analytical layers computed in-house from the same two feeds described above — draws its thesis directly on a live chart, speaks it out loud, states in advance what would prove it wrong, and says so on air when it is wrong.

Concretely, that means the trade stream is stored rather than reduced to candles, so footprint, delta and CVD are reconstructed from actual prints instead of estimated; book snapshots are kept alongside it so replenishment and absorption can be measured rather than assumed; and every reading is stamped with the moment it became knowable, so a level that only makes sense with hindsight never gets drawn as if it had been visible at the time. How it works covers the engine in more detail, and what NAIRO is covers the analyst.

If you want to see order flow read out loud before reading anything else about it, the show streams free on a delay — watch it here. It is educational market analysis, not financial advice, and it issues no signals.