A-Book vs B-Book: Which Side of Your Trade the Broker Takes

A trader who asks whether a broker is A-book or B-book is usually asking one thing: does this firm make money when I lose. The labels bear on that, but they were not coined to answer it, and almost every page explaining them is written for someone buying or operating a brokerage rather than for someone opening an account at one.

What follows treats them as what they are, a description of where the risk of a position goes after the firm accepts it, and separates what a client can verify from what is guesswork.

The One Thing These Labels Describe and the Several They Do Not

A-book and B-book describe the destination of risk. When a firm sends the exposure created by a client position out to an external counterparty, that flow is A-book. When it keeps that exposure on its own balance sheet, the flow is B-book. Everything else attached to the terms was attached by someone else.

They are not the execution-model names a broker advertises. Market maker, STP and ECN describe how prices reach the platform, and how brokers are usually classified follows those names rather than the risk question. A firm can advertise straight-through processing and still retain flow.

They are not a licence category. No regulator issues an A-book or a B-book permission, and what a licence does oblige a firm to do is a separate list altogether. A register records whether an entity may deal on own account, deal as agent, or both, which is related but not the same distinction.

And they are not a quality rating. Neither destination is prohibited, neither is inherently better for a client, and the choice is rarely made once for a whole firm. Flow is sorted by instrument, client and size, so the useful unit is a trade, not a company.

What Happens to Your Order After the Click

Two questions get compressed into one: who the client contracts with, and what that counterparty does with the exposure. Only the second concerns the labels.

The first is settled by the product. A contract for difference and a rolling spot forex position are bilateral agreements between client and provider, concluded away from any exchange. There is no venue to despatch such an order to, so the firm on the other side is the counterparty in every case.

What it does next is where the paths separate. It can enter an offsetting trade with a liquidity provider, passing the risk on. It can hold the exposure. Or it can neutralise part of it against opposing positions already open on its own book.

None of this is visible from the platform. Fill price, timestamp and position line look identical in all three cases, and a delay before a fill says nothing definitive either, because what happens inside the execution window is a price-validation mechanism rather than a signal of where risk ended up. The same applies to why a fill can differ from the price you clicked.

Matching Two Clients Is Not Taking the Other Side

The distinction that decides most of this argument is the one the label hides. A firm holding a client buy order and another client sell order in the same instrument can set them against each other. It then holds no directional exposure: whichever way price moves, one client gains what the other loses and the firm is level.

That is internalisation, filed under B-book because the order never left the firm. It sits in the same bucket as an unhedged position retained deliberately, a genuinely different state. Grouping the two suggests that any order not sent outward is a bet against the client.

Three states are therefore worth separating: exposure passed outward, exposure cancelled against another client, and exposure retained. The first two leave the firm flat. Only the third leaves it holding a position that moves opposite to the client position that created it.

Regulation does not treat matching as a lesser form of dealing. FCA guidance puts back-to-back business, where one client trade is set against another, inside the same category as any other own-account dealing: the firm is a principal there too. Being a principal is therefore not evidence of directional risk.

Where the Broker Makes Its Money in Each Case

The question worth asking is not what a firm calls itself but what its revenue depends on, which follows from the structure of each arrangement rather than from any figure.

Where exposure is passed outward, the firm has bought and sold the same risk. Its income is what it adds to the price obtained, plus any commission, and the client result no longer enters its profit or loss. Where two clients are matched, it collects the difference between the two prices quoted and again holds nothing directional. Both cases reward volume.

Where exposure is retained, the arithmetic changes. The firm holds the mirror of the client position, so the outcome enters its own accounts with the opposite sign. That describes a bilateral contract rather than an accusation: the exposure is normally managed at portfolio level, and can lose as easily as gain.

Read across the bottom two rows and the weakness of the labels shows. Two columns sharing a label share neither risk position nor revenue source, while three columns that differ completely answer the principal question identically.

Dealing on Own Account Is Still Executing Your Order

A common inference runs that if the firm is the counterparty then it is not executing anything on the client behalf, so execution duties fall away. The FCA Handbook states the opposite directly.

Guidance in the best execution chapter says a firm dealing with a client on its own account should be regarded as having executed that order, which brings the chapter into force on the trade, best execution included. Retaining the risk does not move the transaction outside the obligation.

For a retail client the standard has a defined meaning. The best possible result is determined by total consideration, the price of the instrument together with the execution costs directly attributable to it. Price alone is not the measure.

There is also a rule written for this situation. On a product dealt away from a venue, a firm has to satisfy itself that the price it puts to the client is fair, and it has to do so by collecting the market data such a price would be estimated from.

Off a venue there is no public print to measure a fill against, so the reference has to be constructed, and constructing it is the firm obligation.

A further rule closes a gap the labels invite: a firm may not be paid, discounted or otherwise rewarded for sending order flow one way rather than another, where accepting that benefit would breach either the conflicts requirements or the inducement rules.

These are the rules of one regulator, binding the entity it authorises. A group running several entities can offer different protections under each, which is why the entity, not the brand, matters below.

The limiting case is synthetic instruments, where the firm generates the reference price itself, so no external print exists to construct a fair comparison from at all.

What a Broker Has to Tell You, and Why Telling You Is Not Enough

A firm has to build an order execution policy and actually operate it. Per class of instrument, that policy must set out where orders are executed and what decides between one destination and another.

It must then provide appropriate information about that policy, explaining clearly how orders will be executed. Prior consent to the policy is required. And where the policy allows orders to be executed outside a trading venue, the firm must specifically inform clients of that possibility and obtain express prior consent before executing that way.

For a retail forex or CFD account that last provision is the one actually engaged. The consent is normally a clause in the client agreement accepted at account opening, so most traders have already given it without registering what the paragraph was about.

What none of this requires is a label. There is no obligation to describe a policy as A-book or B-book, none to publish the proportion of flow retained, and none to say which state applied to any individual order.

The conflicts rules are where the framing shifts. Every appropriate step has to be taken to spot a conflict between the firm and its client and then either stop it arising or manage it, with a firm’s own pay and incentive arrangements named as one source. Organisational arrangements aimed at keeping conflicts from harming clients must be maintained alongside.

Disclosure enters only after that. It becomes necessary once those arrangements fall short of giving reasonable confidence that clients will be kept from harm, and the disclosure itself has to admit that shortfall in plain terms.

Since 23 October 2025 the Handbook has gone further, requiring a firm to treat disclosure of conflicts as a measure of last resort. Older guidance adds that disclosure does not exempt a firm from maintaining those arrangements, and that over-reliance on it without adequate consideration of how conflicts may be managed is not permitted.

That inverts the usual reading. A conflicts paragraph is commonly presented as the point at which the matter is settled. Under this rule it is a record that prevention was not achieved.

What You Can Verify Before You Deposit

Most attempts to identify a model work backwards from behaviour, reading spreads, requotes or poor fills as evidence. That does not hold: each has causes shared by all three states and none is recorded anywhere a client can audit.

Four things can be checked instead, all of them documents rather than inferences. The first is the legal entity named on the client agreement, often not the brand on the website. The second is that entity on the register of the regulator authorising it, with its permissions, since dealing on own account and dealing as agent are recorded separately.

The third is the order execution policy, which should say whether execution can happen away from a venue, what kinds of counterparty it uses, and whether the firm deals as principal. The fourth is the conflicts of interest disclosure, read in the knowledge of what its presence signifies under the last-resort rule.

Two questions stay out of reach. Which book a particular order landed in is not knowable from outside. Separately, the safety of the money is a different subject, so read what actually protects your deposit rather than treating the execution model as a proxy for it, and checking a broker before you deposit covers the ground around both.

Who This Page Is Not For

It is not for a reader who wants to be told which model to pick. Nothing above ranks them: the destination of risk is not a quality measure, and a firm can serve a client well or badly under any of the three states.

It is not for a reader trying to identify a specific arrangement from platform behaviour, which the middle of this page argues cannot be done.

And it is not a compliance reference. The rules described are those of one regulator, summarised for a trader reading an account agreement, not for a firm designing a policy.

Frequently Asked Questions

What is the difference between an A-book and a B-book broker?

The difference is where the risk of a position goes after the firm accepts it. A-book flow is passed to an external counterparty, so the firm holds no exposure to the outcome. B-book flow stays inside the firm. The terms describe the handling of risk and say nothing about the licence held or the quality of the service.

Does a B-book broker trade against me?

Not necessarily, because B-book covers two situations. If your position is matched against another client going the other way, the firm holds no directional exposure. If it is retained unhedged, the firm does hold a position that moves opposite to yours. Both are called B-book, and nothing visible to you separates them.

Can I tell which book my account is in?

No, not from the platform. Fills, spreads, execution delays and slippage all have causes shared by every model, and flow is normally sorted continuously rather than fixed per account. What can be checked is documentary: the legal entity on the agreement, its permissions on the register, and what the order execution policy says.

Is a hybrid execution model normal?

Yes, and treating a firm as permanently one or the other is the weakest assumption in the discussion. Flow is commonly sorted by instrument, size and client, so one account can produce trades handled in different ways.

Does the execution model change my spread?

Not in a way that identifies it. A quote can be tight or wide under any of the three states, and under FCA rules the best possible result for a retail client is measured by total consideration, meaning price together with the costs directly related to execution.