How Smart Order Routers Pick Venues — and How That's Audited

The invisible decision behind every trade

You tap “buy” on your phone. Somewhere in a data centre, software takes your order, glances at a dozen or so trading venues in the time it takes light to blink, and picks one. If you never think about that moment, you're in good company — most investors don't. But the quality of that split-second choice quietly shapes the price you get.

Let's sit with that over chai, because the machinery is more interesting than it sounds.

First, why choosing a venue is even a question

Decades ago, a stock mostly traded on one exchange. Now the same share can be quoted on multiple exchanges, on multilateral trading facilities (MTFs — European trading platforms that compete with exchanges), in dark pools (venues that don't display quotes before a trade happens), and with systematic internalisers (firms that trade against their own clients directly).

This spread of liquidity across many places is called fragmentation. It arrived largely through rule changes: Regulation NMS in the US in 2005, and MiFID in Europe in 2007, followed by MiFID II in 2018. Both pushed trading away from single national exchanges and toward competition. One source describes US equities as trading across roughly 16 exchanges and more than 30 dark pools — enough venues that a human comparing them by hand would need a very long tea break.

Fragmentation creates a practical problem: the best price to buy might sit on one venue while the best price to sell sits on another. Someone has to check.

What a smart order router actually does

That someone is a smart order router, or SOR: software that automatically scans the venues it's connected to and sends each order, or piece of an order, where the overall outcome looks best right now.

It's worth keeping two ideas separate, because they get muddled constantly. An execution algorithm decides when and how much — slicing a large order over time, perhaps following a benchmark like VWAP (volume-weighted average price) or TWAP (time-weighted average price). The SOR decides where each slice goes. The algorithm handles the schedule; the router handles the venue.

When a router evaluates a venue, price is only the starting point. It weighs fees and rebates (some venues pay you to add liquidity and charge you to take it), how much size is actually available, how likely a fill is, how fast it can reach that venue, and whether sending the order there reveals your intentions to the market. A quoted price of £10.000 on one venue can be more expensive than £10.001 elsewhere once fees are netted out. Good routers compare net cost, not headline price.

Liquidity in dark pools is trickier still, since there are no quotes to look at. Routers lean on historical fill-rate models — and sophisticated ones watch for “toxicity,” meaning prices that tend to move against them after they send an order to a particular pool.

From static tables to adaptive models

The simplest routers follow fixed rules: best price on Venue A, send to Venue A. They're fast, predictable and inflexible. Dynamic routers adjust their weights as they learn which venues fill well for which instruments at which times of day. The most advanced use machine-learning models over many inputs, though those bring a trade-off — they're harder to explain to a regulator and harder to debug. Some firms split the difference: models for scoring, deterministic rules for the final decision.

Speed gets a lot of attention. One vendor describes decision times under 100 microseconds. Whether you need that is a separate question from whether the routing logic is sound.

The uncomfortable truth: “best” isn't guaranteed

Here's the part worth saying plainly. A smart order router improves the odds of a good outcome; it cannot promise the best possible price on any single trade. Quotes move, liquidity disappears, and the price the router saw may be gone by the time the order lands. Nobody can guarantee an execution result, and any claim otherwise deserves scepticism.

Routers also depend on stable conditions. In volatile markets they can become less reliable, which is why human oversight still matters. Implementation is expensive, not every broker offers one, and technical failures happen.

How routing gets audited

Because regulators don't take “trust us” as documentation, routing leaves a trail. Modern routers capture each decision, the market data snapshot behind it, and the path the order took — feeding into transaction cost analysis (TCA) tools that compare fills against what was available.

Jurisdiction matters a great deal here, and the rules are shifting.

In the United States, Rule 606 of Regulation NMS requires brokers to publish quarterly reports on their routing of non-directed orders in NMS stocks and options. Those reports identify the ten venues receiving the most orders, any venue taking 5% or more, and the net payments for order flow, fees and rebates involved — plus a discussion of any arrangement, written or oral, that might influence routing. Customers can also request their own order-level routing information, and brokers that handle substantial not-held order flow must produce more detailed reports on request.

In Europe, the picture has changed. Under MiFID II, firms once published annual RTS 28 reports naming their top five venues, while venues published quarterly RTS 27 execution-quality data. Both are being retired. ESMA issued a public statement in February 2024 saying national regulators should not prioritise supervisory action on RTS 28 reporting while the MiFID II review is transposed into national law — while stressing that the underlying best execution obligation remains. The UK stopped requiring RTS 27 and 28 reports from 1 December 2021.

What replaces the reports is more governance, not less. New European technical standards — Commission Delegated Regulation (EU) 2026/825, in force from August 2026 and applying from 12 February 2028 — repeal RTS 27 and 28 and require firms to maintain detailed order execution policies, a formal register of approved venues, and documented explanations of how routing decisions are made and what weight each factor gets. Firms using automated routing must show those systems operate consistently with best execution obligations. Assessment should use objective data, including consolidated tape data where available. Single-venue firms must evidence that the venue consistently delivers the best possible result. Reviews are required at least annually, with remediation where weaknesses appear.

Notice the direction of travel: from publishing periodic tables that ESMA itself noted were rarely read, toward ongoing, auditable proof. That's a meaningful change in how accountability works — less brochure, more working papers.

What this means if you're not a trading desk

For most retail investors, the router is invisible; your broker's system chooses, and you see the fill. But routing quality affects what you pay, and in the US the Rule 606 reports are publicly posted. They're not light reading, but they do show where orders go and what the broker earns for sending them — useful context when comparing providers. In Europe and the UK, ask your broker what it publishes and how it selects venues.

One gentle observation: an industry that measures time in microseconds still asks clients to read quarterly PDFs. The new European approach, whatever its costs, at least acknowledges that gap.

The honest summary is this: smart order routers are genuinely clever, genuinely useful, and genuinely incapable of guaranteeing perfection. The audit trail exists precisely because everyone involved knows that.