Order routing and execution

Between a submitted order and a confirmed fill sits a sequence of systems, and a fill that arrives in a few tens of milliseconds has passed through all of them. Listing the sequence is worth doing once, because almost every question about execution quality resolves to a question about one specific hop in it.

The order leaves the trading application and crosses the public internet or a leased line to the firm's gateway. The gateway authenticates it and checks it against the account: sufficient margin, a permitted instrument, a valid size, an open market. It then passes to the firm's execution engine, which decides where the order should go. From there it reaches either an internal book or an external counterparty, is matched, and a confirmation travels back along the same path in reverse.

Where the time actually goes 

Not every hop costs the same. The distance between a device and the firm's gateway is usually the largest single component, because it is the only one that crosses an uncontrolled network, and it varies with the connection rather than with anything the firm does. Inside the firm's infrastructure the hops are short and predictable. The hop from the firm to an external venue depends on physical distance, which is why brokers place their servers in the same data centres as the venues they trade with.

The practical reading is that the hop an account holder controls is usually the largest one, and the hops a firm advertises are usually the smallest. A latency figure quoted by any firm measures its own infrastructure from its gateway inward, because that is the only part it can measure or improve.

The two execution models 

Where the order goes after the routing decision is the structural question, and there are two answers. In the first, the firm passes the order to an external liquidity provider or venue, which takes the other side. The firm's own position is flat, its revenue is the commission or the markup on the price it passes through, and its interest is in volume rather than in outcomes. This is usually described as agency or straight through execution.

In the second, the firm takes the other side of the order itself, writing the contract against its own book. Its revenue is the spread plus whatever its book produces, and it manages the resulting exposure by netting client positions against each other and hedging the residual externally. This is usually described as principal or market making execution.

Neither model is inherently better and the industry argument about them is genuinely unsettled. The agency model removes a structural conflict and introduces a dependency on external liquidity, so orders are rejected when that liquidity is absent. The principal model can fill an order the external market would not, and places the firm on the opposite side of the client's result. Most firms operate a hybrid, routing some flow externally and internalising the rest according to rules they publish in outline and not in detail.

Internalisation and netting 

A firm with many clients in the same instrument receives buying and selling interest at the same time, and matching those against each other rather than sending both to an external venue is called internalisation. It is not inherently disadvantageous: an internalised fill avoids crossing an external spread and can be faster, because it never leaves the firm's own systems.

What it does mean is that the firm carries the residual, the imbalance left after client interest has been netted. That residual is what the firm hedges externally, and the size of it relative to total flow is the honest measure of how much market risk a firm is actually running. Firms are not generally required to publish that number, which is why the question of how a firm executes is one worth asking directly rather than inferring from a marketing description.

Last look 

In parts of the currency market, a liquidity provider that receives an order against its quoted price is permitted a brief window in which to accept or reject it. The practice is called last look, and it exists because a provider streaming prices to many recipients cannot update every one of them instantaneously and would otherwise be exposed to being traded against on prices it has already moved away from.

The effect on an order is a short additional delay and a possibility of rejection. Where a rejection occurs, the order returns unfilled and any subsequent attempt meets a market that has moved on. Industry codes of conduct have narrowed the practice considerably, requiring the window to be disclosed and limiting what may be done with the information during it, and some venues have abandoned it entirely. Whether a given order passes through a last look window is a property of the venue it was routed to.

What a best execution obligation actually requires 

Regulated firms operate under an obligation to take sufficient steps to obtain the best possible result for clients, and the obligation is regularly misread as a promise about price. It is not. It is an obligation about a process, and price is one factor in it alongside cost, speed, likelihood of execution and settlement, size and any other relevant consideration. A firm satisfies it by having a documented policy, applying it consistently and monitoring the outcomes it produces.

The practical consequence is that a single fill cannot establish a breach. The obligation is evaluated over a policy and a population of orders, not over an individual one, and that is exactly why execution reporting is done as distributions. It also means the execution policy is a document worth reading, because it states which factors the firm prioritises and for which categories of client and order.

The routing that applies to any given account depends on the platform it runs on, and YAL accounts run on MetaTrader 5, each with its own order handling documented by the platform itself.