A-book and B-book, explained for the person opening the brokerage

Written down that way, B-book sounds like a casino, and trader-facing articles usually leave it there. The practical reason is cost. Offsetting every small trade individually with a liquidity provider costs money on each ticket, and for a book of many small, short-lived retail positions much of that hedging cancels itself out anyway. Carrying that flow internally is often the only economical way to serve small accounts at all, which is why the model is widespread and, in many regulated markets, entirely permitted, provided the conflict of interest is disclosed and managed.

For a founder, the question is which risks you are equipped to carry. A-book converts market risk into a thinner, volume-dependent margin. B-book converts a thin margin into market risk that you must be capitalised and disciplined enough to hold.

The defining requirement is the liquidity relationship. An A-book brokerage needs an account with one or more liquidity providers or a prime-of-prime, which itself is a counterparty onboarding process: due diligence on your entity and licence, a margin deposit held with the provider, and minimum monthly volumes or fees. The provider's terms decide much of your cost base before your first client arrives.

The licence must permit the agency-style activity, which your counsel establishes in your chosen jurisdiction. Operationally you need execution plumbing between your platform and the provider, spread and markup configuration, and monitoring, because your revenue is the difference between the price you receive and the price you show.

The economics are a volume business. A markup measured in fractions of a pip multiplied across monthly volume has to cover everything you run. A-book brokerages therefore live or die on client acquisition and retention, which is why they compete mainly on marketing efficiency and service.

The defining requirement is capital and discipline. A B-book brokerage must be able to absorb a bad month, because there will be one: a news spike, a trending market, or one unusually good client can produce losses that a thinly capitalised desk cannot survive. Jurisdictions that permit dealing on own account generally impose higher capital requirements for this reason, and the specific numbers are a licensing question for your counsel.

Operationally the model demands risk management as a daily practice: exposure limits per client and per instrument, and clear rules for when flow should be hedged externally after all. When a B-book brokerage fails, it is almost always because nobody was watching the book.

The economics are better per client than A-book, which is the temptation. Spread revenue is kept in full and the statistical result of a large, diversified retail book has historically favoured the house. The discipline is remembering that statistics describe large books over long periods, and a small new brokerage is neither.

Mature brokerages rarely run one pure model. The common pattern is a hybrid: flow is carried internally by default, and clients or positions that exceed risk thresholds are routed to liquidity providers. Where each client sits is a risk decision made on measurable behaviour, and the routing rules are among the most commercially sensitive settings a brokerage has.

For a founder the hybrid pattern matters for sequencing. Many new brokerages begin with internally carried flow because the liquidity relationships, deposits and volume commitments of a full A-book setup are heavy for a small book, then add external routing as volume and capital grow. Whether that path is open to you is, again, a function of what your licence permits.

First, what does your licence permit? The models are regulatory categories before they are business strategies, and some licences close one door outright. This is a question for counsel in your jurisdiction and it comes first.

Second, how much risk capital do you have? Count only money that can absorb a losing month without threatening client funds or operations. If the answer is very little, a pure internally carried book is a fragile place to start.

Third, can you get the liquidity relationship? A-book requires a counterparty to accept you, hold your deposit and quote you workable terms as a new entity. That conversation is worth having early, because its outcome may make the decision for you.

Fourth, who is watching the book? A model that keeps risk needs someone whose job is risk. If the founding team has nobody who will own that seat, the choice of model should reflect it.

On the platform side, judge what the software can do today. On the Pelris platform the broker chooses, client by client, between the internally carried model described above and covering the trades with a liquidity provider, the A-book model, and can switch whenever it likes. Brokers bring their own provider or are introduced to one.