Education
A-book, B-book and hybrid: all three are legitimate, opacity is not
What each execution model is, when each one makes sense, why B-book isn't a dirty word, and the single question that separates a serious operator from one that isn't.
There are few topics where the public debate is framed so badly. In broker forums and marketing, "A-book" is presented as virtuous and "B-book" as a scam. The operational reality is a good deal more boring and much more useful: all three models are legitimate, they're used in regulated institutions around the world, and what separates a serious operator from one that isn't isn't which one it uses, but whether it lets you see it.
The three models
A-book: pass-through to the market
In A-book, your client's transaction is hedged in the underlying market. You earn from commission or from a markup on the spread, and you have no directional exposure to the client's result.
Advantage: there's no structural conflict of interest. If your client wins, you win anyway.
Cost: hedging each transaction has a price. On small, highly fragmented flow, hedging costs can eat the margin. And you're exposing all your flow — including the flow that loses systematically — to the market, which is economically suboptimal.
B-book: internalization
In B-book, the transaction is internalized: it isn't hedged externally. The client's result is, in mirror, your result.
Here's the stigma, and it's worth understanding where it comes from. A poorly managed B-book is indeed a problem, because it creates a direct incentive for your client to lose. But a well-managed B-book is simply market making, which is what banks do.
Advantage: you don't pay hedging costs, and you can offer better terms to your client because your margin doesn't depend on an external spread.
Cost: you take on directional risk. If your book is unbalanced and the market moves against you, you pay. This requires a real risk function, not a spreadsheet.
Hybrid: routing by criteria
In hybrid, flow is classified and routed by rules: part goes to A-book, part is internalized.
The usual criteria are client profile, instrument, ticket size, time of day, or historical behavior. A client that trades consistently against the market is hedged; one that doesn't is internalized.
It's the model most serious operations use, precisely because it optimizes cost without taking on indiscriminate directional risk.
Why B-book has a bad reputation (and when it's justified)
The bad reputation comes from a specific and real practice: brokers that internalize all flow, don't manage the resulting risk, and when a client starts to win consistently, degrade its execution or close its account.
That is a scam, and it happens. But the problem isn't internalization: it's the absence of an independent risk function and the lack of transparency about the model applied.
The red flag isn't "they run a B-book." It's "they won't tell me what they run."
The only question that matters
After all of the above, the evaluation comes down to one thing:
Can I see and configure which model applies to my flow?
A serious provider gives you a dashboard where you see, in real time, how your flow is routed, and lets you change the policy without a migration. One that isn't serious dances around the question or gives you a generic answer about "optimized hybrid models."
Follow-up questions that separate a lot:
- Is the configuration by symbol group, by client segment, or global?
- Is changing the split a configuration change or does it require redeployment?
- Is the risk function operationally separated from the trading desk?
- Is there daily reconciliation and who signs it?
- If they use Last-Look, do they publish the rejection rate by upstream counterparty?
The separation of functions, which almost nobody evaluates
This point is technical and it's the one that best predicts whether you'll have problems.
In a well-structured operation, the risk function and the trading desk are operationally separated, with daily reconciliation between them. The reason is simple: if the same person who decides routing is the one who benefits from the B-book result, the conflict of interest isn't theoretical.
When that separation exists, there's a structural barrier — not just a promise — against P&L leaking against the client. When it doesn't, you depend on your counterparty's goodwill, which is a notoriously unreliable control mechanism.
Ask this: is the risk function independent of the desk? How often is it reconciled and who reviews that reconciliation?
How to choose for your operation
If you're a broker just starting out: you probably need hybrid with a conservative configuration. Pure A-book will eat your margin on small flow; pure B-book exposes you to a risk you don't yet have the capacity to manage.
If you have volume and a real risk function: well-calibrated hybrid is where the economics are. You need behavioral data on your clients to classify flow well.
If you run a prop firm: the natural split usually separates evaluation accounts from funded accounts, because they have different risk profiles. They deserve different policies, and you should be able to configure them separately after a single integration.
If you're a fund: you usually want pure A-book and the ability to evidence it. Your allocators will ask, and "trust me" isn't an auditable answer.
What we do
To be concrete: at Exura Prime routing is configurable by relationship, by symbol and by segment, between A-book, B-book and hybrid. Clients see their exposure in real time, with limit and credit controls prior to execution, and the risk function is separated from the desk with daily reconciliation.
Brokers, prop firms and funds onboard as institutional clients and execute through us. Changing the split is a configuration change, not a migration.
If you want the detail by segment: brokers, prop firms and hedge funds. For the full evaluation framework: how a broker chooses liquidity.
Frequently asked questions
Is B-book illegal? No. Internalization is a normal, regulated activity in markets around the world — it's what market makers do. What's regulated is how the conflict of interest is managed and what's disclosed to the client. A broker that internalizes and declares it operates within the rules; one that hides it doesn't.
If my provider runs pure A-book, am I safe from conflicts? From that specific conflict, yes. But pure A-book doesn't protect you from bad execution, opaque markup or a high rejection rate. The risk model is one of the criteria, not the only one.
Can I change models after signing? With a serious provider, yes: it should be a configuration change. If they tell you it requires a migration or a new contract, the infrastructure isn't designed for you to decide.
How do I detect if I'm being internalized without being told? It's hard from the outside, and that's the point. The indirect signals are an abnormally low rejection rate combined with spreads better than the underlying market, or an execution degradation that appears just when an account starts to become profitable. The reliable way to know isn't to detect it: it's to require routing visibility by contract before signing.
