Payment for Order Flow

That is the whole mechanism, and almost every question worth asking about it is about the consequences rather than the description.

Why anyone pays for an order

This is the part most explanations skip, and without it the rest looks like a straightforward kickback.

A market maker profits from the spread between what it pays and what it receives, and loses when it trades against someone who knows something it does not. That second risk — trading with a better-informed counterparty — is the main hazard of the business.

Retail orders are, on average, less likely to be informed than institutional orders. A firm that can separate retail flow from the general pool faces less of that hazard on that flow, and can therefore capture the spread more reliably. That reliability is what it is paying for.

The consequence is that the payment is not a bribe extracted from the customer — it is a share of a genuine economic gain created by separating the flow. Whether the customer receives a fair portion of that gain is the contested question, and it is a different question from whether the gain exists.

Price improvement, and what it is measured against

Orders routed this way are frequently executed at prices better than the prevailing published quote, and brokers point to this as evidence that customers benefit.

The claim is true as stated and the benchmark is the whole argument. Price improvement is measured against the national best bid and offer — the best published quote at that moment. It is not measured against what the order would have achieved under a different market structure, because that outcome does not exist and cannot be observed.

The epistemic centre of the entire dispute, stated as plainly as it can be. Both of these can be true at once, and largely are: a customer receives a price better than the published quote, and the customer might have received a better price still under different arrangements. The first is measurable. The second is a counterfactual and is not. Defenders of the practice cite the first; critics argue about the second. Neither side is being dishonest, and the disagreement persists because the measurement that would settle it is unavailable in principle rather than merely absent. MarketClue does not resolve it. A reader who leaves believing the question is settled in either direction has been told something the evidence does not support.

The conflict, stated without euphemism

The broker chooses where the order goes. One of the things that varies between destinations is how much the broker is paid. That is a conflict of interest, it is not disguised, and it is structural rather than occasional.

The regulatory answer in the United States is disclosure plus a duty of best execution. Best execution requires the broker to seek the most favourable terms reasonably available — which polices the conflict after the fact rather than removing it, and depends on supervision and enforcement to be effective.

Disclosure operates through two rules. Rule 606 requires brokers to publish quarterly reports on where orders were routed and on their payment-for-order-flow arrangements, and to provide order-specific routing information to a customer on request. Rule 605 requires publication of execution-quality statistics.

So the information exists. It is quarterly, aggregated and written for specialists — which is a real limitation and not the same thing as concealment, as What a Brokerage Actually Does sets out about broker revenue generally, and as Slippage and Execution Quality sets out about what the statistics can and cannot settle.

Three jurisdictions have now reached three different answers

This is the most useful thing a reader can know, because it shows the question is genuinely open rather than merely argued.

The European position took two years to become absolute. The prohibition applied directly across the Union from March 2024, but a transitional provision let a member state exempt firms serving its own residents until 30 June 2026. Germany was the only state to invoke it, and when that carve-out lapsed the ban became complete across all twenty-seven members.

Two supervisory clarifications followed and both are worth knowing. In March 2026 the European authorities confirmed that the prohibition applies even where a client has specifically instructed the firm to execute on a particular venue — the rule draws no distinction. And in July 2026 the German regulator published a statement distinguishing business models that remain permissible from arrangements it regards as impermissible circumvention.

Why the natural experiment is less conclusive than it looks

Two large markets now run opposite policies, which appears to offer a clean test. It is worth being careful about that.

The regimes differ in more than one respect — market structure, tick sizes, retail participation rates and venue competition all differ between them, so a difference in outcomes cannot be attributed to the ban alone. And the European ban is recent, with the last exemption having lapsed only weeks before this article was written.

A reader encountering confident claims about the effect of the ban in either direction should note the date, and should apply the standard this portal applies to any short sample, set out in Quantitative Analysis: The Concept and Its Hazards.