Market Quality: Reading Rule 605/606 Reports

Three disclosures, three different questions

Market-structure disclosure exists because **best execution is a duty, not a result** (M9), and a duty you cannot check is a slogan. Three rules do the work, and each answers a different question. **Rule 605** is about execution **quality**. Firms above a size threshold must publish monthly statistics on how customer orders were filled: the quoted spread at the time, the **effective spread** the customer paid, the **price improvement** received, the **fill rate**, and the time to execution. The March 2024 amendments expanded these statistics — more granularity by order type and size, and the threshold lowered to firms with **100,000 or more customer accounts** — with the expanded reports phasing in. Read 605 when your question is *how well was my order filled*. **Rule 606** is about **routing**. Firms must disclose, quarterly, the venues that received their customer orders and the **payment for order flow** they received, broken down by order type and by whether the order was held or routed. Read 606 when your question is *where did my order go, and who was paid*. And the venue side of the same coin — what a wholesaler or exchange itself reports about the quality it delivers — is now Rule **607**. Together they are the only structured evidence that a customer can use to audit the venue the broker chose for them, and the reason a conflict of interest is disclosed rather than forbidden: the model is permitted, and its results are measured. One asymmetry is worth naming at the start. Disclosure covers **customer orders**, which means retail-size orders held by a broker-dealer; it does not describe institutional execution, and it does not describe what the same firm does with proprietary flow. That is not a loophole in the reading, it is the boundary of the data — and knowing the boundary is the difference between using a disclosure and overclaiming from it. Which rule answers which question — How much did my order improve on the quoted price?: Rule 605 — average price improvement · Where did my order go, and was my broker paid?: Rule 606 — routing venues and payment for order flow · What quality does the wholesaler itself deliver?: Rule 607 — market centre disclosures · What did the customer pay against the midpoint?: Rule 605 — the effective spread ←

The metrics, and the identity that ties them together

Four numbers do most of the work, and only one of them is what the customer paid. The **quoted spread** is the market’s, not your broker’s: the best displayed bid and offer, protected by the Order Protection Rule (M8). The **effective spread** is the customer’s: twice the distance between the trade price and the midpoint at the time of the trade, so it captures every improvement and every deterioration relative to fair value. The **realized spread** is the venue’s: twice the distance between the trade price and the midpoint **after** the price has had time to move. And the difference between the last two — effective minus realized — is the **price impact** of the trade, which is the market’s way of saying that someone on the other side knew something. That identity is the whole reading: **effective spread = realized spread + price impact**. A venue that keeps most of the effective spread as realized spread is trading against flow that is not informed — retail flow, mostly — and a venue whose effective spread is nearly all price impact is trading against people who are right. Neither number is a verdict on the customer: the customer’s cost is the effective spread, and the split between realized and impact is a fact about the counterparty. So the professional reading of a 605 table goes in a specific order: compare **effective spread** first, because that is what the customer paid; then look at **price improvement** to see what was returned; then use the **realized** column to understand *why* the venue behaves the way it does; and check **fill rates** because a cheap average fill is worth little if a meaningful share of orders go unfilled. Two further metrics finish the picture. **Price improvement** is measured against the quote, so a firm can report improvement while its effective spread is still wide — improvement is relative to a benchmark that may itself be poor, and the protected quote the benchmark uses does not include everything resting in the market: non-displayed interest and odd lots are not what sets the NBBO, which is why some flow receives improvement that *beats* the display. And **fill rates** must be read with the order mix: a rate of 99.9% means little if the firm only reports marketable orders that had to fill anyway. The intuition to keep: the customer pays the effective spread, the venue earns the realized spread, and the market keeps the difference as the price of being wrong. Every argument about execution quality is really an argument about which of those three should be smaller.

Reading a table honestly

The tables are **averages over different things**. A firm’s 605 numbers blend large and small orders, marketable and resting ones, symbols from penny stocks to megacaps, and different times of day. Two firms can report different averages for the same stock simply because their customer mixes differ — one serving a retail app, the other a community of day traders working limit orders. That makes cross-firm comparison **weak evidence** unless the mix is similar, and it is why the 2024 expansion pushed for more granularity by order type and size: the point of a disclosure is to make like-for-like comparison possible. So use them the way a professional does. First, as a **screen**: a large, stable difference in effective spread between two brokers on the same name is a real question, not noise. Second, as a **trend**: a firm’s own numbers moving over quarters, especially the fill rate or the improvement, is more informative than any single month. Third, as an **audit trail**: with a 606 report you can see whether your orders are all going to one venue, and what that venue pays for them, which is the conflict the disclosure was designed to expose rather than resolve. What the tables cannot do is tell you what would have happened if your broker had routed differently, because the counterfactual is not reported and cannot be. That is the honest limit of the evidence, and it is why the answer to "was I treated well?" is a well-supported inference rather than a proof. The mastery standard for this lesson is being able to say all of that — what the number means, what it is computed from, and which question it cannot answer — while still using it to reach a decision. And because the disclosures describe the current structure, they are also the raw material for the arguments you will have to make in M23: every market-structure debate is ultimately a claim about which of these metrics should move, by how much, and who pays for it. A price-improvement figure is measured against a displayed quote that does not contain the whole market. Improvement that beats the display is real money for the customer and is not by itself evidence that the display was fair — which is why the effective spread, not the improvement, is the number to compare.

Sources

• Execution-quality statistics for customer orders, and the 2024 expansionSEC — Rule 605 and the March 2024 amendments
• Order routing disclosures: venues and payment for order flowSEC — Rule 606 order routing reports
• Wholesaler and venue disclosures of execution qualitySEC — Rule 607 (formerly the order-routing disclosures of market centers)
• Effective spread = realized spread + price impactHuang & Stoll — Dealer versus auction market structures
• Order routing, payment for order flow and retail price improvementFINRA / industry analyses of execution quality and routing