Best execution gets discussed as though it were a pricing question. The conversation usually concerns whether a firm achieved a good level, whether the spread was reasonable, and whether a better price existed somewhere the trader did not look. Those questions matter, but they describe the outcome rather than the obligation. The obligation is evidentiary. A firm has to be able to demonstrate that it took sufficient steps to obtain the best result available to it, under the conditions that existed at the moment of execution, using the information it could reasonably have accessed.
That is a claim about process and about records. It is not a claim about a number.
The distinction becomes uncomfortable quickly in fixed income. Liquidity fragments across dealers, venues, and providers, appears episodically, and prices often emerge through dealer interaction rather than a centralised order book. The counterfactual that best execution analysis depends on, the set of alternatives that genuinely existed at the time, is not sitting in a tape somewhere waiting to be queried. It has to have been captured when it was visible, because it stops being visible almost immediately afterwards.
A great many firms in the small and mid-sized institutional segment produce their execution evidence by reconstruction. The order lives in one system, the quotes were seen in another, the fill was confirmed somewhere else, and the reasoning existed in the judgment of the person who made the decision. When evidence is required, somebody assembles these pieces into a narrative that explains what happened.
The narrative is usually accurate. It is also, in a fairly precise sense, an account rather than a record. It was constructed after the outcome was known, by someone who knows the outcome, from sources that were never designed to be read together. Every element of that description would give an experienced examiner reason to probe further, and the probing tends to focus on exactly the part that reconstruction handles least well, which is the alternatives that were available and not taken.
Consider what the firm actually needs to be able to show. Which liquidity sources were visible at the time. What each of them was showing for that instrument in that size. Why the chosen route was selected over the others. What the trader knew when the decision was made rather than what the market later revealed. A workflow split across an order management system, an execution venue interface, and a spreadsheet can produce most of that with sufficient effort. It cannot easily produce it as a contemporaneous record, and the difference between the two is the entire point of the obligation.
There is an awkward asymmetry developing. Electronic adoption has increased the amount of the market that is observable, which raises what a regulator or a client can reasonably expect a firm to have seen. Around 60% of credit market participants now engage in electronic execution, and high yield electronic trading in the United States moved from 6% in 2015 to 32% in 2025. As more of the market becomes visible in principle, the argument that a particular firm could not have seen it becomes progressively harder to sustain.
At the same time, the visibility is distributed. It sits across more venues and more providers than it did, and a desk observing the market through a traditional order management system may be seeing a smaller proportion of the available liquidity than it was five years ago, even while the total amount of observable liquidity has grown. The standard rises and the firm's view narrows, and those two movements are driven by the same underlying change in market structure.
That is the pattern worth watching. It is not that anyone is behaving badly. It is that the gap between the market that exists and the market a given firm can evidence having examined is widening quietly, and the firms most exposed to it are the ones with the least capacity to notice.
The practical resolution is not more reporting. Bolting an evidence process onto a fragmented workflow produces a second workflow, which the same small team then has to operate alongside the first one, and the additional effort tends to degrade under volume in precisely the periods where the evidence matters most.
The resolution is that the record has to be generated by the act of trading rather than assembled afterwards. If the order carries its intent, its constraints, the liquidity that was visible against it, the route selected, the execution, and the timestamps as a single object, then the evidence exists because the trade happened. Nobody constructs a narrative. Nobody negotiates which system is authoritative. The question of what the trader knew at the moment of decision has an answer that does not depend on anyone's memory.
This is a systems property, not a policy one. A firm can write an admirably rigorous best execution policy and still be unable to evidence compliance with it, because the policy describes what people should do and the systems determine what gets captured while they do it. Where those two diverge, the systems win, and the policy becomes a description of intentions.
Large institutions solve this with transaction cost analysis infrastructure, dedicated surveillance functions, and teams whose job is to hold the evidentiary line. Firms running a handful of seats face an obligation that is not proportionally smaller and a capacity that is dramatically smaller. The regulatory expectation does not scale with headcount.
The response in this segment has often been to accept a somewhat suboptimal position and manage the risk through diligence, which works until it doesn't. It works while volumes are moderate, while the people who remember the decisions are still at the firm, and while nobody asks for a systematic demonstration across a period rather than an explanation of a single trade. Each of those conditions is less reliable than it looks.
There is a more interesting framing available. Evidence quality and execution quality are not separate objectives that compete for the same budget. A system that shows the trader more of the market at the moment of decision is the same system that records what was visible, and the record is a by-product of the visibility rather than an additional cost imposed on top of it. Firms that treat these as one problem tend to spend less than firms that treat them as two.
Rocketfin has been working through exactly this question with desks in the segment, and there will be more to say about it in the near future.
The question ahead of that is one any desk can answer today. If a client asked you to demonstrate, for a single trade last quarter, what liquidity was visible at the moment of execution and why the chosen route was selected, would you be reading a record, or writing one?