Smart order routing: how technology can make order allocation more efficient.

For two decades, order routing in Brazil was a problem solved by the absence of alternatives. There was a trading environment, and the order went there.

This framing has an expiration date. Three projects are advancing simultaneously to compete for fronts currently occupied by a single participant: A5X, focusing on derivatives and anticipated to start operations in the fourth quarter of 2026, along with Base Exchange and CSD BR. It is the first time in over a decade that the concentration of the Brazilian trading market is being contested on multiple fronts at the same time. 

Regulation, for its part, has already anticipated this: since 2022, the CVM Resolution 35 establishes a best execution regime drafted for the case of multiple environments trading the same security.

What is smart order routing?

Smart order routing is the set of systems that decides, in an automated manner, how an order will be executed: in which environment, under what type of offer, in how many parts, at what pace, and at what moment.

Between the client’s click and the closed deal, there is a sequence of decisions that no one sees. An order to buy a thousand shares can become a single offer sent to the book, or thirty smaller offers distributed over twenty minutes, some displayed and others hidden, part executed against the central book and part against the brokerage itself. The client receives an average price and a time. Everything that produced that price happened in milliseconds.

What is the difference between execution algorithm and router?

They are two distinct components, often treated as one. The execution algorithm decides the schedule: it receives the original order and determines how to slice it over time. The router decides the destination: for each generated slice, it chooses the environment and type of offer that provides the best conditions at that moment.

In fragmented markets, both operate in layers. In a concentrated market, the second seems dispensable, but this reading confuses the trading environment with the execution destination.

Why has order routing become a competitive issue for brokerages?

Because brokerage has ceased being the variable of dispute and execution quality has taken its place. When brokerages charged a fee per transaction, competing was simple: it was enough to charge less. That competition ended. With brokerage fees zeroed out in much of retail, the question that no one needed to answer before arose: if the cost to operate is the same everywhere, what distinguishes one brokerage from another?

The answer lies in the price the client actually obtains. Two investors who send the same order at the same moment, through different brokerages, may end up with different prices. The difference is usually a few cents per share. Multiplied by thousands of transactions, it stops being a technical detail and becomes a result.

Now, there are three weighing factors:

Institutional clients measure

Management firms track execution costs by brokerage, by algorithm, and by order size range, reallocating flow based on these numbers.

The obligation is legal, not commercial

CVM Resolution 35 mandates that the intermediary adopts all necessary measures to achieve the best possible outcome, which makes deficient execution a violation of duty, and not just a competitive disadvantage.

The entry of new environments changes the nature of the problem

With more than one possible destination, routing ceases to be configuration and becomes a continuous decision with measurable consequences.

What does regulation require in terms of best execution of orders?

The central duty is in Article 20 of CVM Resolution 35: the intermediary must adopt all necessary measures to achieve the best possible result for the client, taking into account price, cost, speed, likelihood of execution and settlement, volume, nature, and any other relevant considerations.

A rule written before the market existed

It is worth noting when this wording came into effect. It is from 2022, a time when Brazil had only one relevant trading environment for stocks. The CVM wrote a best execution rule for a scenario that did not yet exist. Whoever reads Article 20 today is reading a rule designed for the market that is coming.

Duty of diligence, not of result

The rule requires sufficient measures, not the best price for each isolated transaction. What is evaluated is the adequacy of the process, moving oversight to policy, controls, and produced evidence.

The criterion of total disbursement in retail

The criterion is multifactorial, but with a defined hierarchy for non-qualified investors: in this segment, the rule sets the total disbursement for the operation, made up of the security’s price plus execution costs, as the prevailing factor. Routing that achieves marginally better prices at the cost of higher total costs does not meet the criterion.

Policy, controls, and audit trail

The rule requires the intermediary to maintain rules, procedures, and internal controls capable of obtaining the best available conditions, allowing for the linking at any time between the transmitted order, the respective offer, and the executed deal, and informing clients about the different markets where the security can be traded. The policy needs to explain the factors determining the choice of market when it is not indicated by the client.

The linking requirement has a direct consequence on architecture. When a parent order is sliced into dozens of child orders, with cancellations and replacements, the ability to reconstruct the complete chain needs to be anticipated in the system design, not resolved later by log cross-referencing.

The European model adopted by the CVM

The duty falls on the intermediary, who is responsible for deciding the execution market. There is no order protection rule mandating automatic routing to the displayed best price, as occurs in the United States under Regulation NMS. Compliance is proven through policy, execution consistent with it, and documented evidence.

Does smart order routing make sense in a concentrated market like Brazil?

It does, as long as one understands what the routing decision means when there is a single central book. The variable is not the environment, but the path to execution.

The objection is reasonable and arises frequently: if every order goes to B3, what exactly is there to route? The answer is that “going to B3” does not describe a decision; it describes an address. Within that address, there are choices with a direct effect on the price obtained, and one of them doesn’t even involve the central book.

RLP: the routing decision that already exists

The RLP offer allows the intermediary to act as a counterparty to retail client orders. The price is continuously adjusted by the trading platform, and the deal always closes at a price equal to or better than that of the central book. When the spread of the book is equal to or greater than two tick sizes, the intermediary has an obligation to improve the price by at least one tick. Offers from other clients of the same intermediary at the best price in the book take priority.

This mechanism is exclusive to aggressive orders from retail individuals, with optional adherence, and applies to mini dollar contracts and index contracts, as well as to stocks selected by B3. Deciding when to activate RLP and when to send to the central book is materially a routing decision.

Type of offer, timing, and large lots

The other three decisions are the type of offer — market, limit, with hidden or minimum quantity, tied to the best price —, timing, between opening auction, continuous session, and closing auction, and the treatment of large lots, whose minimum lots the CVM has taken responsibility for disclosing, with the stated goal of avoiding unwanted effects on price formation in the central book.

The architecture argument

There is a legitimate counterpoint: building a complete multi-environment router for stocks today means developing capabilities for which there is no alternative destination. The answer is not to anticipate the entire construction, but to avoid an irreversible decision. Execution systems often treat the trading environment as a constant embedded in code, rather than as a parameter. When the second environment comes into operation, the difference between the two approaches will not be marginal effort, but rewriting.

For Jorge Sellmer, Chief Revenue Officer (CRO) of Objective, the standard is not exclusive to the capital market. The study Data Maturity, conducted by Objective in 2025, classified companies into five levels — beginner, explorer, structured, data-driven, and data-driven — and found significant concentration in the early stages, with fragmented data use and decisions still supported by intuition.

Applied to order execution, the diagnosis translates directly. A brokerage that routes by a parameter fixed years ago, without measuring the outcome obtained, remains in the early stages of maturity, regardless of the sophistication of its infrastructure. What separates one level from another is not the technology installed, but the ability to know if it is working.

The risk of rewriting is not hypothetical. In a multinational credit analysis company, Objective participated in the migration of an AS400 and COBOL environment — the central platform of the main business lines, with decades of accumulated rules — to a cloud microservices architecture, without interrupting operations. The acceleration by AI agents increased the squad’s delivery by 37% and, in the same movement, multiplied defects per sprint until process governance reached the new speed. The lesson is transferable: technical capacity without a measurement method does not produce maturity; it produces faster instability. Read the full case.

How to divide large orders without impacting the price?

There is no execution of large orders without impact. There is impact management and a conscious choice of the point on a trade-off curve.

Why do large orders move the price?

The reason is arithmetic. The order book has limited depth. An order that consumes all the available volume at the best price starts executing at the second-best, then the third. Those selling to that order perceive the demand and adjust the price upwards. In the end, the executor paid more just because they wanted to buy a lot.

Slicing solves part of the problem and creates another. Executing slowly reduces the instantaneous consumption of liquidity but keeps the order exposed to the market longer, which can move against it for reasons unrelated to the transaction. The practical question is not how to eliminate the impact, but how long it is worth waiting to reduce it.

Temporary impact and permanent impact

The temporary impact reflects the consumption of liquidity and tends to dissipate after execution. The permanent impact reflects the information that the negotiation reveals to the market and is not reversible.

Microstructure literature establishes that impact grows approximately with the square root of the ratio between the size of the order and the volume traded in the period. The implication is direct: doubling the size of the order does not double the cost, and reducing participation in volume has diminishing returns. There exists a point beyond which extending execution no longer compensates.

Slicing strategies

How not to signal intent

Slicing solves half the problem. The other half is not revealing what you intend to do. Displaying only part of the offer reduces the information available to the market, at the cost of lower priority on the hidden portion. Regular schedules and identical sizes produce recognizable signatures, which makes controlled variations in size and interval a protection, rather than a detail. The ratio between passive offers, which capture the spread but risk non-execution, and aggressive offers, which ensure execution at the cost of the spread, is the main parameter of aggressiveness in the strategy.

How to measure if routing is truly delivering better execution?

Through systematic transaction cost analysis, with explicit references and breakdown of deviation.

Average price does not answer anything

Was an order executed at R$ 32.40 well or poorly executed? The question cannot be answered without knowing what price the security was at when the order arrived, how much the market moved in between, and what happened to the price immediately afterward. Without reference, the number does not inform anything.

Choice of reference

The most commonly used is the arrival price, as it allows for the decomposition of total cost. The VWAP for the interval is common but is manipulable by the order itself when it is large relative to the market. The closing price only makes sense when the mandate uses that reference. The choice is not technical; it is contractual: it needs to correspond to what the client considers as the outcome.

Decomposition of implementation shortfall

The deviation from the arrival price decomposes into four parts: cost of delay, market impact, time cost, and opportunity cost of the unexecuted portion.

Decomposition is what makes the number actionable because the same aggregated deviation can have causes with opposite solutions. High delay indicates an infrastructure or internal queue problem. High impact indicates a strategy too aggressive for the size. High opportunity cost indicates the opposite.

Execution quality indicators

Aggregated averages hide the cases that matter. The cuts that often produce discoveries are by order size range, by asset liquidity, by volatility regime, and by algorithm parameter versions.

Post-trade reversal: the test that validates price improvement

When execution occurs at an apparently favorable price and the market immediately returns to the previous level, the gain measured at the time of the transaction does not hold. It is the appropriate measure to evaluate internalization mechanisms, including RLP: the question is not whether the price was equal to or better than the book at that moment, but whether the execution would have been better in the central book a few seconds later.

The bridge to compliance

The same data supports the demonstration of compliance with Article 20. A institution capable of showing the distribution of total disbursement of its retail clients, segmented by routing decision, is not just optimizing results: it is producing evidence that the best execution policy operates as declared.

Final considerations

Order routing combines three attributes that rarely coincide: it is an express regulatory obligation, it is a measurable competitive differential, and it is an engineering problem with a known technical solution. The window to treat it as an architectural project, rather than a response to a market event, is the interval between now and the entry of new environments into operation.

Objective operates in building decision systems in regulated, high-volume environments, integrating software engineering, data, and artificial intelligence for financial institutions and the capital market, focusing on low-latency processing, compliance automation, and modernization of critical platforms. Talk to a specialist.

Frequently asked questions about the topic