Designing challenge into the compliance workflow

Why second-line review needs structured context, clear accountability and an auditable conclusion.

Why second-line review needs structured context, clear accountability and an auditable conclusion

Most of the attention on RTS 6 assurance lands on the self-assessment itself: the document, the conclusion, the evidence behind it. Less attention goes to a function the regulation assumes runs underneath the whole framework, and in places requires explicitly: second-line review.

It is tempting to treat "second line" as a governance label rather than a working requirement: a box on an org chart occupied by Risk Management or Compliance, sitting somewhere between the trading desk and Internal Audit. RTS 6 does not treat it that way. Where the regulation addresses second-line review directly, it is specific about what the function has to be able to do, not just where it has to sit.

Two provisions make that explicit.

Article 9 hands primary responsibility for the annual self-assessment to a firm's risk management function, with Internal Audit reviewing where applicable and senior management providing approval. That is a second-line function being given direct ownership of the firm's central RTS 6 conclusion, not a role reviewing someone else's.

Article 16(2) goes further. Real-time monitoring of algorithmic trading activity has to run on two lines of defence: the trader running the algorithm, and a risk management or independent risk control function. ESMA's February 2026 supervisory briefing spells out what independence means here. The second-line function cannot be hierarchically dependent on the trader, cannot be directly involved in the trader's own activity, and needs what the briefing calls the powers, tools and procedures to challenge the trader. ESMA is explicit about what does not qualify: a head of trading desk is not independent from that desk's own traders, and an IT function, however close to the technology, is not considered equipped to challenge one independently.

That framing is useful because it clarifies what second-line review is supposed to produce. Not a second signature. Challenge: genuine, informed, capable of changing an outcome. And challenge of that kind depends on three things being true, none of which follows automatically from a reviewer simply sitting in the right seat.

Structured context turns independence into genuine challenge

A reviewer can be hierarchically independent, formally empowered, and still unable to challenge anything meaningfully, if the information reaching them does not support it.

The FCA's August 2025 review of algorithmic trading controls found this gap directly. It reported that the technical knowledge of algorithmic trading among compliance staff, and the level of oversight and challenge compliance actually provided, varied significantly between the firms it reviewed. In some firms, compliance had strong technical grounding and provided real challenge to trading processes. In others, weaker technical knowledge limited the ability to challenge trading behaviour, and compliance was less involved in the processes it was meant to be reviewing.

That is not simply a training gap. A reviewer who only sees a conclusion has nothing to test it against, whether the conclusion is that the self-assessment is compliant, that an algorithm was approved, or that a pre-trade control passed. Genuine challenge needs the scope behind the conclusion, the control that was relied on, and evidence that the control actually operated as intended, not just that it exists on paper.

This is where fragmentation does real damage. When the material a second-line reviewer needs sits across a self-assessment spreadsheet, a policy on SharePoint, calibration rationale in a slide deck, and control-trigger statistics in a system only risk logs into, review quietly narrows to whether the box was ticked. Not because anyone lowered the bar deliberately, but because assembling the full picture every time is too costly to do consistently.

ESMA's own guidance on pre-trade controls shows the scale of what a reviewer should actually be able to see: the calibration methodology and its rationale, the quantitative data behind each threshold, and statistics on how often a control has actually triggered, all of it meant to feed back into whether the control still works. A reviewer working from a document that only states a control exists is in no position to challenge whether it still functions.

The same gap shows up over time, not only at a point in time. ESMA has flagged that a series of small recalibrations can accumulate into a materially different algorithm without any single change looking significant enough to trigger review on its own. Looking at one change in isolation cannot show that pattern. A connected view of what changed, when, and why, can.

Clear accountability gives that challenge somewhere to go

The second requirement is closely related: a reviewer needs someone to challenge, not just something.

The FCA's review found this gap sitting in some of the most basic records a firm keeps. Good practice, it noted, included algorithmic inventories with a clear indication of who owned each algorithm and who was approved to operate it. In some firms, though, the inventory did not name the individuals approved to operate a given algorithm at all. The same section of the review raised a related point: in some cases, compliance staff had limited oversight of pre-trade and post-trade controls, leaving them with a weak understanding of how those controls actually worked.

Put those findings together and the second line's position becomes clear. If nobody is named against a control, there is nobody specific to question, follow up with, or hold to a deadline. Review becomes a comment on a document rather than a conversation with an owner.

RTS 6 treats this as more than administrative tidiness. Article 9 requires firms to assess, as one of four distinct elements of the annual self-assessment, their governance, accountability and approval framework specifically, alongside the algorithms themselves, business continuity, and overall compliance. Accountability is not incidental to the assessment; it is one of the things being assessed. The same expectation runs through ESMA's guidance on pre-trade controls, which requires that any revision to the parameters underpinning a hard block involve risk management and compliance directly, and that traders cannot override those parameters on their own. The rule assumes a named second-line party is in the room whenever a control changes, not that one was consulted somewhere along the way.

An auditable conclusion makes the second line's own view defensible

The first two requirements make challenge possible. The third is about what happens to that challenge afterwards.

Second-line review reaches conclusions too: satisfied, escalated, remediation required. Each of those conclusions is itself an assertion, in the same way the underlying self-assessment is an assertion. It needs the same thing: a record of what was examined, what was queried, what was not accepted at face value, and what follow-up was required before sign-off. Without that, the second line's own conclusion is no more defensible than the first-line conclusion it was meant to test.

This matters for two audiences in particular. Internal Audit, in reviewing the self-assessment, is not only checking the first line's work. It is implicitly checking whether the second line's challenge was real. A file showing only an approval, with no trace of what was probed to reach it, gives Internal Audit nothing to assess. And a regulator asking why a firm reached a particular conclusion on a specific control or algorithm is, in substance, asking the second line the same question the firm's self-assessment has to answer for RTS 6 as a whole: not just what was concluded, but why.

A second-line review can be thought of as its own short chain:

Control → Rationale → Evidence of operation → Challenge raised → Response received → Conclusion

When each link is captured as review happens, the conclusion carries its own reasoning with it. When it isn't, the conclusion is only as strong as the recollection of whoever reached it. And recollection has a habit of leaving the firm before the question gets asked.

Three requirements, one function

None of this calls for more process. It calls for treating second-line review as what RTS 6 already assumes it to be: a function that only works if it has enough context to form a genuine view, enough clarity about ownership to direct its challenge somewhere specific, and enough of a record to make its own conclusion stand up later.

Strip out any one of the three and the other two lose most of their value. Context without accountability tells a reviewer something is wrong but not who should fix it. Accountability without context gives a reviewer someone to ask, but nothing informed to ask them. Challenge without a record, however real at the time, leaves a firm no better placed to answer why second-line review reached its conclusion than it is to answer the same question about the self-assessment itself.

The FCA and ESMA are not asking firms to build something new here. Independence, accountability and validation are already written into RTS 6. What both regulators are increasingly testing is whether firms can show it. That is a much harder thing to demonstrate from a spreadsheet, an inbox and an org chart than it is to assert in a policy document.

Daniel Shearer, Founder of Cerylis

Daniel Shearer is the founder of Cerylis and a capital-markets compliance specialist with experience across investment banking, professional services and RegTech. His work has covered MiFID II and MiFIR, algorithmic and electronic trading governance, RTS 6, regulatory controls and wider markets compliance.