Risk & Dealing8 min read

A-Book, B-Book and Hybrid: Choosing How Your Brokerage Handles Orders

What each execution model means, what it pays, what it risks and how real brokerages combine them.

BridgexFX editorial · 2026-08-22

Fast forex trade execution on trading screens

The A-book versus B-book argument is usually conducted at full volume and with very little detail. One camp treats internalizing flow as gambling, the other treats pure STP as the only respectable option, and most brokerages quietly operate somewhere between the two. Underneath the noise sits an ordinary risk decision with costs on each side. Here is a clear look at all three models.

Three models in plain terms

With an A-book (STP) model, the client's order goes straight to a liquidity provider. The broker is paid through markup or commission and holds no market risk on the trade, so it earns from volume rather than from clients losing.

With a B-book model, the broker takes the other side of the client's trade itself. Its results are the mirror of its clients': it gains when they lose and is exposed when they win.

A hybrid model sorts flow. Orders from clients who look informed or toxic are sent out to the market, while flow that looks uninformed is kept, with real-time rules making the call.

What each one pays and what it risks

A-book is clean and predictable, but revenue is just markup multiplied by volume, so you need meaningful scale to cover fixed costs. It also has the simplest story to tell regulators and clients, because there is no conflict of interest to explain.

B-book can be very profitable because most retail flow loses over time. The catch is concentration: one undetected winner, or a gap through stops, can wipe out months of income quickly. It is a risk business that looks like a revenue business until it doesn't.

Hybrid is where most of the industry sits. When segmentation is good, it offers the best risk-adjusted return. When segmentation is crude, it combines the weaknesses of both.

Segmentation is the whole game

A hybrid broker lives or dies by how well it classifies clients. The warning signs are well known: latency arbitrage patterns, trading into news spikes, steady short-term profitability, HFT-style order-to-trade ratios and groups of correlated accounts. Detecting them in real time is mature technology.

What goes wrong is upkeep. Traders adapt, so rules need constant tuning, and they need human oversight too. Fully automated classification punishes good clients with false positives, while fully manual review can't keep up. The brokers who manage this well review their rules weekly instead of setting them once.

Regulators and reputation

Best execution, conflict-of-interest disclosure and order-handling transparency requirements are getting stricter across jurisdictions. Whatever you run, document it, disclose what you must, and make sure your marketing never promises something your execution contradicts.

There is a reputational upside to openness. Several of the better-known brokers now describe their hybrid model in their legal documents and even their advertising. In an industry with a trust gap, explaining exactly how orders are handled can be a selling point.

Where a new broker should begin

New brokers are tempted to B-book everything for the margin and then discover the risk work they skipped. A safer starting point is a conservative hybrid: A-book by default for anything your system flags, B-book only for clearly uninformed flow, and tight exposure limits per client and per symbol.

As data accumulates, your segmentation improves and you can internalize more with confidence. Let the model follow your data, not your ambition. And whichever route you take, a dealing desk, in-house or outsourced, is not optional. An unmonitored B-book is just an unhedged bet.

The takeaway

A-book is clean volume income, B-book is profitable but volatile, and a well-run hybrid is where most serious brokerages end up. Pick the model your risk team can actually operate, disclose it honestly and make sure your marketing never promises more than your execution delivers.

#Execution models#Hedging#Flow segmentation
Blog

Did this article spark an idea?

Let's scope it together, honestly and in writing.