Execution splits are where a brokerage turns its dealing policy into live market behavior. Knowing how to configure execution splits means more than assigning a percentage of volume to A-Book and retaining the rest internally. It means deciding which flow should reach external liquidity, which flow can be warehoused, and when those decisions must change based on measurable risk.
For Forex and CFD brokers, a split configuration affects P&L volatility, liquidity cost, fill quality, client experience, and operational exposure at the same time. A static 70/30 rule may be easy to explain, but it rarely reflects how different instruments, client segments, market sessions, and trade sizes behave. The objective is controlled flexibility: a routing model that can protect the book without introducing avoidable latency, inconsistent fills, or a reliance on engineering tickets.
Start with the commercial purpose of each split
Before building routing logic, define what the split is intended to achieve. The same execution split can be appropriate for one brokerage and dangerous for another depending on capitalization, liquidity relationships, client acquisition channels, and risk appetite.
A new broker may use higher externalization rates while its risk team builds confidence in client behavior and operational controls. An established broker with a diversified retail flow may retain more eligible exposure internally to reduce liquidity costs and monetize balanced client flow. Neither approach is automatically better. The relevant question is whether the retained risk can be monitored, hedged, and funded under stressed market conditions.
Execution splits should therefore be attached to a clear policy. Define the instruments covered, the account groups affected, maximum net exposure, allowed hold time, escalation thresholds, and the person or team authorized to change the rule. If a rule cannot be explained to risk, dealing, compliance, and finance in plain language, it is not ready for production.
Build the routing hierarchy before assigning percentages
A percentage split is only one layer of the configuration. A stronger model begins with a routing hierarchy that establishes which conditions override the default allocation.
For example, a brokerage might begin with a 50/50 split for a standard EUR/USD retail group. But the execution flow should first check whether the order exceeds a defined notional threshold, whether current net exposure is close to its limit, whether the client belongs to a restricted segment, and whether external liquidity is currently available at acceptable depth. These conditions should take priority over the default percentage.
This prevents a common operational mistake: treating all orders as equivalent. A $1,000 trade from a low-frequency client and a concentrated $500,000 order during a major data release do not create the same risk, even when they are on the same symbol.
A practical hierarchy usually evaluates four areas:
- Client classification, including account type, region, trading history, profitability profile, and internal risk designation.
- Instrument behavior, including volatility, liquidity depth, session risk, spread stability, and gap exposure.
- Order characteristics, including notional size, direction, leverage, holding period, and frequency.
- Book conditions, including current net exposure, concentration, hedge capacity, and external venue health.
Each layer should produce an observable decision. Risk teams need to see why an order was internalized, split, delayed, or passed through, not simply that it received an execution status.
Separate default splits from hard risk limits
A default split is a commercial routing preference. A hard risk limit is a control that cannot be overridden by normal order flow. Keep these concepts separate.
For instance, a broker may route 40% of eligible XAU/USD volume externally under normal conditions. That is a default. If internal gold exposure reaches a predefined dollar limit, all additional exposure in the same direction may need to be hedged externally. That is a hard limit.
Blending both into one percentage rule creates ambiguity. During fast markets, ambiguity becomes manual intervention, and manual intervention is where response time and consistency deteriorate. Hard limits should automatically override routing percentages and generate an immediate alert for the dealing desk.
Configure segments around behavior, not assumptions
The most useful execution splits are based on observed flow, not broad labels such as “retail” or “professional.” Those labels may be relevant for compliance and suitability, but they do not reliably predict execution risk.
Start with a manageable number of segments. A typical framework may distinguish new accounts, established low-frequency traders, high-turnover traders, news-sensitive flow, large-notional accounts, and accounts with a verified history of adverse selection. Over-segmentation creates maintenance work and can make a routing policy impossible to audit. Under-segmentation leaves the broker applying the same economics to materially different behavior.
The data used for segmentation should be specific. Review win rate, average holding time, trade frequency, instrument concentration, slippage sensitivity, behavior around macroeconomic releases, and directional correlation across accounts. A trader who profits consistently from short-duration trades around liquidity events may deserve different routing treatment than a long-horizon client with similar monthly volume.
Segmentation must also be reviewed for fairness and regulatory alignment. The purpose is risk management and execution control, not arbitrary discrimination or manipulation of client outcomes. Routing logic should never be used to intentionally degrade fills, delay valid orders without justification, or selectively reject profitable customers.
Use instrument-level rules for volatile products
A single global split does not account for the different microstructures of major FX pairs, exotic currencies, indices, metals, equities, and crypto CFDs. Each product can create a different combination of liquidity cost, overnight gap risk, client concentration, and hedge availability.
Highly liquid majors may support tighter controls and faster external hedging. Gold and indices often require more conservative exposure caps around market opens and macroeconomic events. Crypto CFDs may need tighter notional controls because price dislocations can move faster than a broker can rebalance exposure. Exotic FX pairs can have thin depth even when quoted spreads appear acceptable.
Set splits at the symbol or instrument-group level, then apply session and event overlays where necessary. An otherwise acceptable internalization rate at 2:00 p.m. may be unsuitable minutes before a central bank decision. The goal is not to eliminate internalization. It is to avoid carrying exposure when price formation and hedge costs are least predictable.
Test execution splits against real order flow
Do not deploy a new split configuration based solely on a commercial target. Test it against historical orders and, where possible, run it in a controlled environment before it affects live routing.
The testing question is not simply whether the broker would have made more money. Examine how the rule would have changed net exposure, hedge frequency, external execution costs, slippage, rejected orders, and drawdown during normal and stressed periods. A configuration that improves average revenue but creates unacceptable tail risk is not a successful design.
Use representative periods that include quiet sessions, volatile sessions, major news events, and periods of unusual client activity. Model the effect of latency and partial fills as well. A split that looks efficient in a spreadsheet may fail when an external liquidity provider widens, rejects, or provides less depth than expected.
After deployment, compare actual results against the model. If externalization is consistently more expensive than expected, investigate liquidity quality and routing logic. If retained exposure is growing faster than modeled, review client segmentation and limit thresholds. Configuration is an operating process, not a one-time setup.
Make real-time visibility part of the design
Execution splits lose value when the team can only review them after the close. Dealing desks need real-time visibility into exposure by symbol, client group, direction, and liquidity source. They also need to see the current routing path for each order and the reason a rule was triggered.
This is where programmable execution infrastructure matters. In ZeroMS, brokers can build visual execution flows for A-Book, B-Book, splits, and delays, then monitor outcomes as conditions change. The practical advantage is operational control: authorized teams can adjust logic without waiting for custom bridge development, while retaining an auditable view of the decision path.
Set alerts around meaningful conditions rather than generating noise. Useful alerts include exposure approaching a hard limit, a sharp increase in rejected hedges, deterioration in fill latency, repeated routing overrides, or an unexpected concentration of flow in one instrument. Every alert should have an owner and a defined action. Visibility without accountability is only a dashboard.
Review the configuration after market changes
A split that performs well in stable conditions may become unsuitable after a new affiliate campaign, a change in leverage, the addition of a liquidity provider, or a shift in the product mix. Review routing performance on a regular schedule and immediately after material operational changes.
Focus on exceptions. Which accounts repeatedly trigger overrides? Which symbols produce the largest gap between modeled and actual hedge cost? Which liquidity sources fail during peak demand? These answers identify whether the problem is the percentage, the segmentation, the risk limit, or the external execution path.
The best execution split is not the one with the highest internalization rate. It is the one that gives the brokerage a repeatable balance of client execution quality, controlled market risk, and predictable unit economics. Build that balance into the rule design, measure it continuously, and give the dealing desk the authority to act before a small exposure becomes a larger operational problem.