Learn order flow trading: the complete map
Most order flow material hands you four instruments and leaves you to guess which to open first. This page maps the discipline instead: what each family measures, the sequence that makes every stage checkable before the next, and where the map runs out.
Order flow reads two records: executions — what traded at each price and which side crossed the spread — and resting orders, the size waiting to be traded. Its four tool families are volume profile, footprint, depth of market and liquidity heatmap. Learn them in that order: each stage makes the next one readable.
What order flow is, before any tool
Two datasets sit underneath every instrument in this field, and confusing them is the most expensive mistake in the subject. The first is the record of executions: completed transactions, each with a price, a size, a timestamp and — on venues that publish it — the side that crossed the spread to get filled. The second is the book of resting orders: limit orders waiting at each price, which have not traded and can be withdrawn without notice. Executions are receipts. Resting size is an advertisement.
A candle discards both. Open, high, low, close: four numbers standing in for thousands of transactions and every cancellation between them. Nothing there says whether buyers had to pay up or were handed their fills, whether the high was refused by size or simply ran out of buyers, or whether anything was waiting where price turned. Every order flow tool recovers one slice of what that compression threw away.
The mechanism worth learning before any of them fits in a sentence. In a central limit order book, a price changes for exactly two reasons: aggression consumes all the resting size at the best level, or that size is withdrawn. So a large trade can print with no movement at all — a 250-lot market buy arriving into 400 resting at the offer is filled in full, leaves 150 sitting there, and the price does not move a tick. The tape records something worth knowing; the chart records nothing. Learning to see that gap is the whole subject in miniature, and it costs nothing.
The four families, and what each one is made of
Every tool you will meet is a rearrangement of those two datasets. Two families are built from executions and two from the resting book, and that split predicts almost everything about how far each one can be trusted. Execution-based tools describe events that are finished and cannot be taken back. Book-based tools describe a state that can evaporate in a millisecond, and frequently does.
Read the last column first. Each family is defined as much by its blind spot as by its measurement, and the blind spots are complementary by design: a volume profile knows where business happened but not who was aggressive, which is exactly what a footprint chart adds. The tape — the raw stream of prints, before anything aggregates it — sits underneath both.
A fifth thing is sometimes sold as a family of its own: market by order data, which decomposes each price level into individual orders with identifiers and queue positions. It is a higher resolution of the same book, not a different measurement — a specialised instrument to justify later, never a starting requirement.
| Family | Built from | What it measures | What it cannot see |
|---|---|---|---|
| Volume profile | Executions, bucketed by price | Where business accumulated over a window | Which side was aggressive, and in what order |
| Footprint chart | Executions tagged bid or ask | Which side crossed the spread, price by price | Orders that were placed but never traded |
| Depth of market | Resting orders, at this instant | The size currently advertised at each price | Hidden size, and whether it will still be there |
| Liquidity heatmap | Resting orders, kept over time | Where displayed size was placed, held or pulled | Whether a band was consumed or simply cancelled |
The order to learn them in, and why that order
The sequence is not a matter of taste, and it is not the order the tools are usually sold in. Three properties decide it. Reviewability: some of this data survives the session, some exists only in the instant. Dependency: a reading of aggression means nothing until it is attached to a location. Cost: each step demands more of the feed, and the expensive tiers cannot be judged until the cheap ones are understood.
The strongest argument for the sequence is reviewability. A profile is still there after the close; a reading you got wrong on a Tuesday can be examined on a Saturday. A ladder is gone the instant it changes, so a mistake made there is unfalsifiable in practice. Learning in the order of what can be reviewed is learning in the order of what can be corrected.
The second argument is that each stage supplies the missing input of the next. Structure without executions gives levels and no way to tell whether they are defended. Executions without structure give readings with nowhere to put them: heavy selling in the middle of a range is a curiosity, the same selling at the edge of yesterday's value area is a question worth asking. The book without either gives numbers changing faster than they can be read, and no reason to care about any of them.
- The mechanism, with no tool at all. How a match happens, why price moves, what aggressive and passive fills are, and how a large trade can leave no trace on a chart. This can be worked through on paper, and it is the only stage that never goes stale.
- Structure: where business was done. Volume profile and Market Profile, the point of control, the value area, the difference between a price the market accepted and one it only passed through. First among the tools because it produces locations, and every later reading is a question asked at a location.
- Executions: who paid to get filled. The tape first, then the footprint and delta built on top of it. This is where the two readings that matter live: absorption, where aggression meets larger passive size and dies against it, and exhaustion, where the aggression simply stops arriving.
- The live book: what is waiting. The depth of market ladder and the liquidity heatmap. Deliberately last, because displayed size is the least reliable object in the discipline and the most seductive one on the screen.
- Your own record. A written sample of your own readings with the outcomes attached. Without it you are collecting vocabulary, and vocabulary accumulates whether or not it is working.
How to know a stage is finished
"Learn order flow" has no completion signal, which is why it is possible to cycle through the same material for years while feeling that something is progressing. Replace the feeling with a test: each check below can be attempted in writing, away from a screen, and either passes or does not.
None of them are difficult. They are specific, which is what the usual self-assessment lacks: I understand absorption cannot fail, and a claim that cannot fail teaches nothing.
- Mechanism. Explain, without drawing anything, how a large trade can print with no price change, and how price can move several ticks on almost no volume. Two mechanisms, two sentences.
- Structure. Mark the levels you would work around before the session opens, in writing, and state for each one the observation that would make it irrelevant. If nothing could, it is not a level, it is a decoration.
- Executions. Take a bar with a large one-sided delta and state what would have to happen next for it to be absorption rather than the start of a break — before knowing which it was. The point is committing in advance, because both readings are available afterwards.
- The book. Name three reasons a displayed order can disappear, only one of which is that somebody traded against it. Anyone who cannot is about to learn it from a ladder instead, which is the expensive way.
- Your record. State which of your own readings you get wrong most often, and give the count from your own notes. If there is no number, the stage is not finished, however long it has taken.
What order flow cannot do
This is the section that decides whether the rest of the page is worth anything. Order flow has hard limits, and none of them are removed by a better tool, a faster feed or more screen time. Anyone presenting the discipline as a way of removing uncertainty is selling the thing this page exists to argue against. The honest summary is one line long: measurement narrows the question, it does not answer it.
- It does not predict direction. It describes what has already been executed; the last minute constrains the next one loosely at best, and most readings are compatible with the move simply continuing.
- It does not tell you when. Absorption at a level can hold for one bar or for an afternoon, and nothing in the print distinguishes those two cases at the moment it appears.
- It cannot see identity or intent. No feed available to a retail participant labels a counterparty. You can observe that unusual size traded at a price; attributing it to a category of participant is an interpretation laid on top, not something read out of the data.
- It cannot see what was never displayed. Stop orders are not resting in the book, iceberg orders show a fraction of their size, and some intentions sit in somebody's head. The measurement is real; it is a measurement of the visible part.
- It is only as good as the feed carrying it. Aggregated, delayed or venue-fragmented data changes the numbers being read. Where trading is split across many venues, the flow on your screen may be a minority of the whole, and the aggressor side inferred rather than published.
- Almost every observation has at least two readings. Heavy selling into a bid is absorption if price holds and the beginning of a break if it does not. The tooling cannot tell you which, and no future version of it will.
Where the learning usually goes wrong
Four failure modes account for most of the time people lose in this subject, and the sequence above is built around them.
- Starting at the ladder. It is the most animated thing on the screen and the least dependable. Displayed size is free to place and free to revoke; the executed columns beside it are the only part describing something that actually happened.
- Collecting tools instead of questions. A second instrument is not a second opinion. If the rule is act when either one agrees, nothing has been filtered — the number of qualifying prices has been doubled. Decide in advance which tool picks the location and which one holds the veto.
- Treating a reading as a signal. A large print, a one-sided delta, a wall on the ladder: each is an observation with at least two interpretations. The discipline tells you where to pay attention and whether a level is being defended. The decision still needs a context and an invalidation written before entry.
- Letting the settings drift. Row size on a profile, the imbalance ratio on a footprint, where the session is cut: these are not cosmetics, they are part of the claim. Two traders with different settings read different levels on identical data and neither is wrong. Fix yours, write them down, and never compare two readings built under different ones.
How to start without spending anything
The mechanism stage costs nothing, and it is the one that most changes how a chart looks afterwards. Matching, aggression, passive fills, queue position, why a cancellation moves a price without a single contract trading — all of it can be worked through with a pen.
The structure stage needs only a series of volume by price, the cheapest order flow data there is. Build profiles on sessions that have already closed — and prefer them to live ones for a reason unrelated to money: you can stop, write the reading down, and only then look at what followed. Live, you cannot, and the temptation is to remember the readings that worked.
Write the reading before the outcome. Two sentences is enough: what you think is happening, and what would show you that you were wrong. A dated folder of those is worth more than any instrument you could add — it is the only artefact on this page that measures you rather than the market.
When a term here is unfamiliar, the glossary defines each one with a worked example and the mistake it causes. A route through it: order flow trading for the frame, auction market theory for why any of it works at all, then the four families in the order above.
In short
- Order flow is built from two records that behave differently: executions, which are finished, and resting orders, which can vanish before they trade.
- The four tool families split two and two along that line — profile and footprint from executions, ladder and heatmap from the book.
- Learn them in order of reviewability: what survives the close before what exists only in the instant, because only the first kind lets you check yourself.
- Each stage supplies the missing input of the next. Structure gives locations, executions say whether a location is defended, the book says what is advertised there.
- None of it predicts direction, says when, or sees intent — and none of it contains position sizing, which sits outside the measurement entirely.
Frequently asked
- How long does it take to learn order flow trading?
- No honest answer exists as a duration, and any figure offered is a marketing number rather than a measurement. What can be said is what determines it: how much of the mechanism was understood before a tool was opened, whether sessions are reviewed after the close or only watched live, and whether readings are written down with their outcomes attached. The rest is counted in reviewed sessions, not in weeks.
- Which order flow tool should I learn first?
- The volume profile, because it produces locations and every other reading is a question asked at a location. It also survives the close, so a reading can be checked long afterwards. The depth of market ladder is the usual first choice and the worst one: the most animated object on the screen, the least dependable, and the only one that cannot be reviewed.
- Do I need Level 2 or market by order data to start?
- No. Footprint charts, delta and the tape are built from executions rather than from the resting book, and they need no order identifiers at all. Market by order answers a narrower question about how a price level is composed and how its queue moves, and it is the most expensive tier in every stack. It is something to justify after the cheaper stages, not a prerequisite for them.
- Can you learn order flow on stocks or crypto?
- The mechanism travels. Anywhere a central limit order book matches buyers and sellers continuously, the same relationship between aggression and displacement holds. Data quality travels less well: some venues publish the aggressor side and others leave it to be inferred from where a trade printed, and trading split across many venues means the flow on your screen may be a fraction of the whole. Learn the mechanism anywhere; judge the numbers by what the venue actually publishes.
- Is order flow the same as volume analysis?
- Volume analysis is part of it, not all of it. A volume profile aggregates total contracts at each price and takes no side at all. Order flow adds two things a volume series alone cannot give: which participant crossed the spread to get filled, and what size was resting and waiting there. Colouring a plain volume histogram by buyers and sellers is a category error, and some tools do it anyway.