What order flow actually tells you — and what it doesn’t

Most retail tools show you a candle. A candle is a summary written after the fact: four numbers describing what happened over a period that was chosen for you. Order flow is what the candle was built out of — the raw sequence of trades that executed, in the order they executed, with the size of each one.

That difference matters for one reason. By the time a five-minute candle closes green with heavy volume, the buying that made it green is finished. The people who acted on it were watching the trades, not the candle.

What a print actually contains

Every executed trade — a print — carries four facts and no more:

• The pair. Which market it happened on. SOL/USDT and SOL/BTC are different books with different participants.
• The side. Whether the trade was buyer-initiated or seller-initiated. This is the taker rule: if someone crossed the spread to buy, the print is marked a buy.
• The size. How much changed hands.
• The time. Down to the millisecond.

That is the entire factual content. Everything else you hear about order flow — accumulation, distribution, absorption, smart money — is interpretation layered on top of those four fields. The interpretations can be useful. They are not data.

Volume is not direction

The single most common mistake is reading a volume spike as a directional signal. It is not. A spike says the same amount of buying and selling changed hands, because every trade has two sides. What varies is who was impatient — who crossed the spread instead of waiting.

That asymmetry is real information, and it is what the buy/sell split measures. But even a lopsided split does not mean price will follow. A wall of aggressive buying that gets absorbed without the price moving is a bearish observation, not a bullish one: it means someone was willing to sell everything that eagerness could pay for.

A signal says money moved. It does not say the price will follow. Anyone selling you the second claim is selling you something they cannot deliver.

Size means nothing until you normalise it

A 200,000 USDT trade is a rounding error on BTC/USDT and a violent event on a small-cap pair. Any alert built on a raw currency figure will therefore be simultaneously too loud on the majors and completely deaf on everything else — which is exactly how people end up muting alerts within a week.

There are two sane ways to normalise. The first is to express size in a common asset — for example, measuring every trade in bitcoin terms, so one threshold behaves consistently across pairs. The second is to express size relative to that pair’s own recent average: “three times the median trade on this book over the last hour”. Both beat a flat dollar number. We wrote separately about why thresholds belong in BTC.

What order flow cannot tell you

It is worth being blunt about the limits, because most content on this subject is not.

• Intent. A large buy might be accumulation, a short cover, a market-maker hedging an options position, or an index fund rebalancing. The print looks identical in all four cases.
• Who. Exchange trade feeds are anonymous. Any tool claiming to identify “institutional” flow from a public feed is inferring, not observing.
• What happens next. Flow is a description of the past few seconds. It has predictive value only in the weak, statistical sense that unusual activity tends to cluster.
• The other venues. One exchange is one venue. Large participants deliberately split orders across several, and over the counter entirely.

Using it without staring at a screen

The practical value of order flow is not that it predicts. It is that it is early — it happens before the chart pattern that summarises it. That makes it a good trigger for attention and a bad trigger for execution.

A workflow that survives contact with real life looks roughly like this:

1. Define what counts as unusual on the coins you actually care about, in normalised terms.
2. Let something watch for it so you do not have to. A rule that fires a handful of times a day is useful; a rule that fires forty times is a live feed you will learn to ignore.
3. When it fires, look. Open the pair, check whether the flow continues or gets absorbed, check whether the same thing is happening on related pairs.
4. Make the decision on your own criteria. The alert did its job the moment it made you look at the right chart at the right time.

That is the whole discipline. The tooling exists to compress step 1 and 2 so the human spends attention only on steps 3 and 4.

Common questions

Written by the team building Pharos Trade. We make an alerting tool, not a trading service — nothing here is investment advice, and none of it predicts a price. Figures in examples are illustrative. Last reviewed 2026-08-23.