Evidence-led broker guideBroker execution and risk models

Brokerage Models Compared: A-Book vs B-Book vs Hybrid

Compare A-book, B-book and hybrid brokerage models by counterparty, market risk, revenue, conflicts and the technology each needs for routing and reporting.

By Orrnn Traders2,915 words
Broker order-routing architecture with risk checks, internal and external execution paths, exception handling, and reconciliation
Execution design is a controlled workflow: intake, validation, routing, exception handling, counterparty processing, and reconciliation.

Direct answer

A brokerage model describes where a client's trade-related market risk goes. In an A-book model, the broker passes client flow or the resulting exposure to an external liquidity provider or venue. In a B-book model, the broker keeps some or all of that exposure. A hybrid model handles different orders, clients or products in different ways.

Key takeaways

  • A-book, B-book and hybrid describe where client trade risk goes. They are shorthand, not legal definitions.
  • The client contract, the real order and hedge flow, and the entity that bears risk matter more than a vendor's label.
  • Each model changes conflict controls, capital planning and technology: routing, exposure monitoring, rules and best-execution evidence.
  • Retail CFD rules on leverage, margin close-out and benefits apply by product and jurisdiction, whichever model is used.
  • Before accepting any model claim, ask for a real order traced through price, routing, exposure and audit logs.
  • Model choice does not settle the licence you need. The regulated activity and jurisdiction do, so take qualified legal advice.

Define the three brokerage models

Each label answers one question: where does the risk from a client's trade go, and who is the counterparty?

A-book usually describes arrangements where client flow or the resulting exposure is passed to an external liquidity provider or venue. The client usually contracts with the broker, while the hedge or routed order goes to an external counterparty. B-book usually describes a broker that keeps and internalizes some or all of the exposure, so the broker's own result moves against client gains and losses. Hybrid can mean routing different orders, clients, products or risk amounts differently, or hedging only aggregate exposure after internalization.

Begin with facts: who is the client's contractual counterparty; whether an order is executed against the broker, agency-routed, or handled under another structure; when and how any hedge occurs; how the client price is formed; what discretion exists; and which entity bears market, credit, operational, and conduct risk. Obtain legal advice for the relevant product, client type, and jurisdiction.

Do not market the label as a guarantee. External routing does not guarantee a fill, a particular spread, no conflict, or good execution. Internalization does not by itself prove manipulation or poor treatment. The assessable questions are whether the model is permitted, accurately disclosed, governed, adequately capitalized, operationally controlled, and producing outcomes consistent with applicable obligations and client agreements.

Brokerage models at a glance

The table compares the three brokerage models on the questions that drive contracts, controls and system design.

Read each row as a default pattern, not a fixed rule. A brokerage can run more than one model, and the actual contract, product, jurisdiction and configuration decide the answer. The figure shows the decision points where client orders, routing and hedge flows separate. Each path needs its own record.

Flow diagram in which a client order reaches a broker decision point and then takes an A-book path to an external liquidity provider, a B-book path where the broker keeps the exposure, or a hybrid rule that splits between them, with a record kept for each path.
Routing flow for A-book, B-book and hybrid treatment. The diagram marks the decisions a brokerage must record; it does not prescribe a model or describe any supplier's configuration. Open full-size diagram
Brokerage models at a glance
ModelCounterpartyWhere market risk sitsRevenue sourceConflict-of-interest controlsTechnology implication
A-bookClient contracts with the broker; the hedge or routed order goes to an external liquidity provider or venueHedged portion passed to the external counterparty; the broker keeps timing, slippage and counterparty riskSpread or markup and commissions, less external execution costDocumented routing rules, price-source checks, and evidence that counterparties are chosen for stated reasonsLiquidity bridge or FIX connectivity, routing logs, and reconciliation of external executions to client trades
B-bookClient contracts with the broker, which takes the other side of the trade as principalMainly with the broker, which keeps some or all of the exposureSpread or markup, commissions and client losses on retained positions, which are uncertain and not a predictable profitIndependent risk limits, dealing-desk governance, remuneration review, and disclosure matched to the actual modelExposure engine with enforced limits, price controls, and records of every dealing decision
HybridClient contracts with the broker; orders or exposure go to more than one route or counterparty under rulesSplit by rule, product, client group, amount, or aggregate hedge policyA combination of the above, plus any fees that depend on the route usedVersioned routing rules, independent review of rule changes, and outcome monitoring by client groupRules engine, multi-venue connectivity, exposure aggregation, and reporting that reconstructs each decision

Map orders, trades, positions, and hedges separately

Draw the client order lifecycle from request through validation, price, acceptance, execution, position, cash posting, correction, and reporting. Then draw the broker's risk lifecycle: exposure creation, aggregation, limit check, hedge decision, external order, execution, residual risk, financing, and reconciliation. A client execution and a broker hedge may be separate events with different timestamps, prices, and counterparties.

Record the identifiers linking client orders, executions, internal positions, hedge baskets, external orders, counterparty executions, and ledger entries. If several client trades produce one hedge, the allocation and performance analysis should remain reproducible. If a hedge is delayed or sized by net exposure, document the approved logic rather than retrospectively describing every client order as directly transmitted.

State the systems of record for client order state, external execution, exposure, limits, and accounting. Reconcile quantities, prices, fees, currencies, financing, and timestamps. Disagreements are expected during timeouts, partial fills, corrections, and disconnections; define uncertainty states and escalation instead of forcing premature closure.

Questions the brokerage model label cannot answer
TopicEvidence neededWhy it matters
Legal capacityClient agreement and legal analysisDetermines rights, disclosures, and applicable duties
Price formationSources, aggregation, markups, timestampsExplains the price presented and executed
RoutingVersioned rule, inputs, outcome, overrideShows how discretion and automation operate
HedgingExposure policy, limits, hedge recordsSeparates client execution from risk transfer
ConflictsGovernance, remuneration, monitoringTests whether incentives can harm client outcomes
Execution qualityOrder-level and aggregate measurementsReplaces marketing claims with outcomes

Understand economics without reducing the model to spreads

Revenue and risk can include spread or markup, commissions, financing, conversion, platform or data fees, partner remuneration, external execution cost, market movement on retained exposure, hedging slippage, rejected hedges, credit exposure, and operational losses. Model gross and net economics by product and scenario. Do not describe client losses as predictable brokerage profit; exposure, withdrawal, hedging, conduct, concentration, and reputation make outcomes uncertain.

Externalization introduces counterparty spreads, commissions, financing, minimums, margin, credit, market-data, connectivity, and operational cost. Internalization can reduce some external costs but increases retained market and concentration risk and can amplify conflicts. Hybrid models add routing and surveillance complexity. Compare capital usage, liquidity under stress, and downside, not only expected unit margin.

Product rules also limit the economics. ESMA reminds firms that products meeting the definition of a CFD are subject to applicable national product-intervention measures, including leverage limits, a mandatory risk warning, margin close-out, negative balance protection, and a prohibition on monetary and non-monetary benefits. A model whose economics depend on client losses, or on benefits paid by a counterparty, needs legal review against these limits in each jurisdiction.

Build severe-but-plausible scenarios: one-way client exposure, gap moves, correlated instruments, stale prices, counterparty failure, rejected hedges, widened spreads, market closure, deposit or withdrawal stress, fraud, system outage, and manual dealing error. Include the timing of collateral calls and cash settlement. Board-approved appetite should translate into pre-trade and intraday limits that the technology can enforce.

Govern routing, dealing, and conflicts

Routing rules should use permitted, documented inputs and produce an auditable outcome. Define client or account grouping, product, size, exposure, counterparty availability, cost, market state, limits, fallback, rejection, and override. Review whether grouping could create unfair or discriminatory outcomes. Store rule versions and reconstruct the decision for a historical order.

Separate commercial targets from control decisions. Dealer remuneration should not encourage harmful rejections, delay, price asymmetry, or selective treatment. Establish independent compliance and risk oversight, access segregation, change approval, surveillance, exception reports, and escalation. Manual intervention should require a reason and produce a record, not disappear into an administrator log.

Conflict disclosures must be specific enough for the applicable regime and actual model, but disclosure alone may not cure a conflict. In the UK, the FCA Handbook's conflicts-of-interest chapter (SYSC 10) is the starting point for how firms identify and manage conflicts. Legal advisers should assess governance, consent, best-execution or fair-dealing duties, product rules, and prohibited practices. Technology teams should implement the approved policy; they should not invent it through defaults.

Measure execution outcomes across comparable groups

Capture request, quote, acceptance, routing, acknowledgement, execution, rejection, cancellation, and client-notification timestamps with known clock quality. Retain available market context, requested and executed quantity, price, slippage, partial fills, reason codes, and counterparty. Define how requotes, last look where applicable, price improvement, and asymmetric slippage are represented.

Analyze distributions rather than one average. Segment by product, order type, side, size, session, volatility, client group, route, counterparty, device or region where relevant. Review rejection, timeout, fill rate, partial fill, latency, realized spread, and price comparison using a defensible benchmark. Small samples and market differences can mislead, so state methodology and confidence limitations.

Connect execution monitoring to complaints and incidents. A cluster of adverse outcomes may result from a mapping defect, stale source, capacity problem, counterparty behavior, routing rule, or market conditions. Investigation should preserve raw data and trace configuration changes. Correct affected records and client outcomes under an approved process where required, then verify that the root cause is removed.

Technology implications of each broker model

Each model moves authority between systems. Orrnn does not supply liquidity or hold a brokerage licence, so these responsibilities must be assigned in the contract.

Liquidity routing and bridge. An A-book needs stable external connectivity, counterparty adapters, and a record of each hedge or route. A B-book needs less external routing but still needs a price feed and a written rule for when the broker's own prices differ from external sources. A hybrid needs both, plus a rules engine that states which path each order takes. Confirm the supported counterparties, dialects and FIX versions for the proposed build.

Risk engine and exposure monitoring. A B-book depends on current, complete positions, limits that move with price and intraday alerts that the risk team can act on. An A-book still needs exposure measurement to decide when and how much to hedge, and it needs a view of residual risk when a hedge fails. Price and routing services need stale-data checks, source health, approved markups and safe degradation. A risk dashboard fed by the same failed source as the dealing system gives little independent assurance.

Rules engine and change control. Routing and dealing rules should be data with versions, owners and approval records, not hidden defaults. Store the rule version active at each order time so a historical decision can be reconstructed. Changes to grouping, thresholds or counterparty order should follow the same approval path as a product change.

Reporting and best-execution evidence. Whatever the model, the firm needs order-level timestamps, quotes, routing decisions, executions and client notifications, plus aggregate reports by client group, product and counterparty. Where MiFID II applies, the obligations sit with the investment firm, not with the software; the platform has to produce the evidence the firm needs to meet them.

Design operations for uncertain orders and hedges

Order and hedge workflows must handle duplicates, late messages, sequence gaps, partial fills, contradictory states, and reconnects. Maintain distinct client and external order state machines linked by stable identifiers. Do not assume a timeout means no execution. Queue uncertain cases, limit further risk if necessary, and reconcile with independent records or counterparty confirmation.

Operational runbooks should cover counterparty outage, bridge failure, platform disconnection, fast market, instrument suspension, erroneous price, limit breach, manual hedge, correction, and end-of-day break. Name decision authority and client-communication responsibilities. Rehearse scenarios using realistic records and verify accounting after recovery.

Evaluate vendors without outsourcing the model

A trading platform, bridge, risk engine, or liquidity provider may support several models, but a capability does not approve its use. Require documentation of price, order, routing, risk, audit, and reporting behavior. Demonstrate representative flows and negative cases. Identify which decisions remain with the brokerage and which service provider personnel can change configuration or intervene.

Do not accept claims such as 'pure A-book', 'conflict free', 'no dealing desk', or 'best price' without definitions and evidence. Ask whether the statement describes the client contract, each order, net exposure, a default configuration, or a marketing category. Obtain independent legal review and verify outcomes from production-quality data.

The RTX5 roadmap lists its liquidity bridge and FIX 4.4 and 5.0 adapters as build not published (checked 9 November 2026), and a price-list inclusion does not specify an execution policy. Ask the demonstration to trace a proposed A-book, B-book or hybrid rule through price formation, order routing, exposure updates and audit logs. Then record the supported model and each party's responsibilities in the contract.

Decision criteria

Choose the model after permissions and product scope are known, not before. A label chosen for marketing reasons is not a control design.

Use the checklist below as a gate before a model is approved for a launch or a material change. It does not replace legal advice, a regulator register check, or a tested demonstration. This guide does not provide legal, tax or investment advice, and it does not show that any brokerage model is profitable, lawful for every client, or safer for traders.

Decision checklist

  • Legal capacity, products, clients, entities and disclosures match the actual order and risk flow
  • Client executions and brokerage hedges are recorded as distinct events with linked identifiers
  • Risk appetite is translated into enforced exposure, credit and concentration limits
  • Routing and intervention rules are versioned, approved, monitored and reconstructable
  • Conflicts and remuneration receive independent compliance and board oversight
  • Execution monitoring uses order-level evidence and comparable client cohorts
  • Counterparty, order, position, cash and fee records reconcile on a defined schedule
  • Stress, outage, error and uncertain-order procedures have been rehearsed with realistic records

Frequently asked questions

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

An A-book broker passes client trades or the resulting exposure to an external liquidity provider or venue, so market risk is largely hedged or routed outside the firm. A B-book broker keeps some or all of that exposure, so its own result depends on client outcomes. Neither label is a legal definition. The contract, the actual hedge flow and the entity that bears the risk decide what the model means in practice.

Does a B-book model always mean the broker trades against its clients?

Not automatically, but it creates a conflict that must be controlled. When the broker keeps client exposure, its profit can move against a client's gain. Regulators look at how that conflict is identified, governed and disclosed, and at client outcomes over time. Internalization alone does not prove unfair treatment, and external routing does not guarantee good execution. The broker must still show the controls and records behind its model.

Is a hybrid brokerage model safer than pure A-book or B-book?

Not by default. A hybrid model can spread risk, but it adds routing rules, more counterparties and more places where a rule can behave unexpectedly. The added complexity needs versioned rules, monitoring by client group and reconciliation across each path. Compare capital use, liquidity under stress and downside outcomes for each model. Do not treat the hybrid label as a safety claim.

What technology does each brokerage model need?

An A-book model needs reliable external connectivity, counterparty adapters and hedge records. A B-book model needs real-time exposure, enforced limits and dealing-desk audit trails. A hybrid model needs both, plus a rules engine that records which path each order took. All models need order-level timestamps, reconciliation and reports that support best-execution and conflict reviews. The exact scope depends on the contracted products and the firm's responsibilities.

Does choosing an A-book or B-book model change the licence a broker needs?

The model is one factor, but the licence or permission depends on the regulated activity, the products, the client type and the jurisdiction. A broker may need authorization for dealing or arranging activities whichever model it uses. A technology vendor does not supply a licence. Obtain jurisdiction-specific legal advice before choosing a model, and check the current regulator register for the exact legal entity.

Primary sources and evidence boundary

Sources are listed to support specific definitions, public vendor statements, and regulatory frameworks. They do not endorse Orrnn, prove that a product meets a requirement, or replace a current proposal, contract, legal opinion, technical test, or regulator decision.

The original source review was completed on 21 September 2026. RTX5 pricing and plan references were updated on 26 September 2026; the dated note below identifies that product source. Recheck time-sensitive requirements and commercial terms before relying on them.

  1. RTX5 broker pricing and plan inclusions

    RTX5 / Orrnn

    Product pricing reviewed 26 September 2026. Vendor-published commercial scope, not independent proof of performance, compatibility or regulatory approval. Inclusions vary by plan.

  2. Markets in Financial Instruments Directive (MiFID II)

    EUR-Lex

    Official EU directive text; national implementation and the firm's facts determine applicability.

  3. MiFID II

    European Securities and Markets Authority

    Official ESMA materials for the EU regime, including conduct and execution-related rules where in scope.

  4. ESMA reminds firms of their obligations under CFD product-intervention measures

    European Securities and Markets Authority

    Official ESMA statement that CFDs meeting the definition are subject to national leverage limits, risk warnings, margin close-out, negative balance protection and benefit restrictions. National rules and the product's classification must be checked.

  5. FCA Handbook: SYSC 10 Conflicts of interest

    Financial Conduct Authority

    Official FCA Handbook chapter on conflicts of interest. Applicability depends on the firm's permissions and status; counsel must confirm the rules that apply.

  6. FIX Protocol

    FIX Trading Community

    Primary technical source for standardized trading messages; it does not define a brokerage's legal model or guarantee interoperability.

  7. Intermediary Registration

    Commodity Futures Trading Commission

    Official U.S. intermediary overview illustrating why activities and products, not marketing labels, drive analysis.

Your next step

Explore an RTX5 configuration for your business.

Compare the published platform plans, then share the account capacity, client workflows and connectivity your business needs. A requirements discussion can establish the demonstration scope, quote and implementation dependencies. Module inclusions vary by plan.

Platform plans and inclusions