Case Study 2: Finding Gaps in the Bond Tape

Rafael Torres and the Non-Equity Post-Trade Reporting Remediation

Overview

Client institution: Farida Asset Management (fictional), a Paris-based EU investment firm managing EUR 12 billion in multi-asset credit strategies, including investment grade and high-yield corporate bonds, structured credit, and sovereign bonds Engagement lead: Rafael Torres (independent regulatory consultant, formerly VP Compliance Technology at Meridian Capital) Support: Priya Nair (Big 4 RegTech consultant, junior member of the engagement team) Regulatory trigger: FCA/AMF dual-jurisdiction review finding gaps in bond post-trade transparency reporting; AMF (Autorité des marchés financiers) supervisory letter citing MiFIR Article 21 deficiencies Primary challenge: Accurately classifying bonds as liquid or illiquid under ESMA FIRDS, and building an operational post-trade reporting process robust enough to handle the heterogeneous nature of the EU bond universe Outcome: Full remediation of APA reporting infrastructure over four months; firm avoided enforcement action

Background: Why Bond Transparency Is Hard

When Rafael Torres first encountered MiFID II's non-equity transparency requirements in his years at Meridian Capital, his reaction was the same as most compliance professionals who had come up through equity markets: the framework looked similar to equity post-trade transparency, with a liquid/illiquid overlay that seemed relatively straightforward to implement.

It was not straightforward.

The equity world has a simple transparency logic: most equities are liquid, publication windows are short, and the ISIN-based classification is well-established. The bond world is different in almost every dimension:

• The global bond universe contains hundreds of thousands of ISINs — far more than the equity universe.
• The overwhelming majority of bond ISINs are illiquid by any reasonable standard. AFME estimated that as of 2022, more than 98% of EU corporate bond ISINs classified under ESMA FIRDS were illiquid.
• Bond trading is predominantly OTC — bilateral negotiation between an investment firm and a dealer — rather than exchange-based. This means the investment firm itself bears the APA reporting obligation under MiFIR Article 21, not the venue.
• Bond ISIN structures are complex: the same issuer may have dozens of different bonds outstanding, each with different maturity, coupon, currency, and seniority — and therefore potentially different liquidity classifications.
• FIRDS classifications change quarterly. An instrument classified as illiquid in Q1 may become liquid in Q2 if its trading volume increases — or vice versa.

For Farida Asset Management, all of these complexities had converged into an operational reporting failure that the AMF had identified during a routine supervisory review.

The Regulatory Trigger

The AMF supervisory review had been triggered by an ESMA data quality report comparing post-trade transparency data published by APAs against transaction report data submitted by ARMs. ESMA's data analysis team had noted an anomaly: Farida's ARM transaction reports for OTC bond trades showed significantly higher volume and trade counts than Farida's APA post-trade transparency submissions for the same instruments and the same period.

In simple terms: Farida was telling the regulator about its bond trades (ARM), but not telling the market (APA). The gap was substantial — approximately 35% of OTC bond trades in the review period had ARM transaction records but no corresponding APA publication.

The AMF issued a supervisory letter (not a public enforcement notice — at this stage, it was a private regulatory inquiry) requesting an explanation of the discrepancy and a remediation plan within 60 days.

Farida's chief compliance officer called Rafael.

Initial Assessment: Why Was the Gap So Large?

Rafael and Priya spent the first two weeks on-site conducting a root-cause analysis. They interviewed Farida's operations team, reviewed the post-trade reporting process documentation, and analyzed a sample of trades in the discrepancy period.

They found four root causes, each feeding the others.

Root Cause 1: Confusing APA reporting with ARM reporting

Farida's operations team had built its post-trade workflow around the ARM transaction reporting obligation — specifically, the end-of-day T+1 transaction report. The ARM reporting was largely correct: the firm's middleware vendor handled ARM submissions reliably, and the AMF confirmed that transaction reports were being received.

However, the team had not built a separate workflow for APA post-trade transparency. They had been under the impression — reinforced by informal guidance from a previous operations manager who had since left the firm — that ARM transaction reporting satisfied both the regulatory surveillance obligation and the market transparency obligation.

This was incorrect, and Rafael had seen it before. "The APA obligation and the ARM obligation are parallel tracks," he told Farida's CCO. "They come from the same trade event, but they go to different recipients for different purposes. ARM goes to the regulator, confidentially. APA goes to the public. You were meeting the regulator's need. You were not meeting the market's need. MiFIR requires both."

Root Cause 2: The FIRDS classification problem

For the minority of bond trades that Farida was reporting to an APA, the operations team had built a simple binary logic: if the ISIN was in a spreadsheet maintained by the head of fixed income operations (a spreadsheet last updated in 2022), it was classified as liquid; otherwise it was illiquid. Liquid bonds were reported within 15 minutes; illiquid bonds were batched for end-of-day submission.

This classification system had three critical flaws:

First, the spreadsheet was 18 months out of date. ESMA publishes FIRDS updates quarterly, meaning that the firm's classification could be wrong by up to six quarters of classification changes.

Second, the spreadsheet covered only bonds that Farida had traded in the prior year. New ISINs — new bond issuances, instruments from newly accessed credit markets — were defaulted to "illiquid" without any FIRDS lookup, which was conservative but meant the firm was potentially delaying publication for liquid instruments.

Third, the spreadsheet contained classification errors — bonds that ESMA had classified as liquid were listed as illiquid, and vice versa. This was not the result of intentional mislabelling but of the difficulty of maintaining accurate FIRDS data manually across a universe of hundreds of ISINs.

Root Cause 3: OTC bilateral trades falling through the reporting gap

Farida's APA connectivity was configured only for trades executed on trading venues (regulated markets, MTFs, OTFs). The system automatically generated APA reports for venue-executed trades. But OTC bilateral bond trades — negotiated directly between Farida and a dealer, executed outside any venue — required a manual trigger in the reporting system.

The manual trigger had been implemented when the firm first built its post-trade reporting infrastructure in 2018. Over time, as trading volumes grew and the operations team turned over, the manual process degraded. Some trade types — particularly block trades in investment grade credit, and structured credit transactions — were not consistently triggering the manual APA report.

These were precisely the trades that appeared in the ARM data but not the APA data. OTC bilateral trades, by definition, do not appear on any venue's trade feed. The ARM obligation was captured because the ARM middleware was fed from the trade capture system. The APA obligation was not captured because it required manual action that was not being taken.

Root Cause 4: No reconciliation between ARM and APA submissions

If Cornerstone's case study (Chapter 20, Case Study 1) was about a near-miss in equity dark pool monitoring, Farida's case was about the absence of a basic reconciliation control. There was no automated or even manual process that compared the list of trades reported to the ARM against the list of trades submitted to the APA. If such a reconciliation had existed and run daily, the discrepancy would have been visible from the first week of the reporting gap.

The reconciliation gap was, in Rafael's view, the most preventable failure of all. "You can debate whether FIRDS classification is hard — it is hard," he said. "But the reconciliation logic is simple: every trade in the ARM feed should have a corresponding APA submission, or a documented reason why it doesn't. If that check runs daily, you catch the gap in 24 hours, not in an AMF supervisory letter."

The FIRDS Deep Dive: Why Classification Is Technically Challenging

Priya led the FIRDS classification work, and it became the most technically demanding part of the engagement.

ESMA's FIRDS database is publicly available at registers.esma.europa.eu. It contains reference data for instruments admitted to trading on EU venues, including liquidity classification data published under RTS 2. The challenge is that accessing and interpreting this data at scale is non-trivial.

Volume: FIRDS contains reference data for millions of ISINs across all asset classes. Downloading and parsing the full dataset requires infrastructure capable of handling large XML or CSV files.

Update frequency: FIRDS is updated quarterly for the standard liquidity determination. However, for newly issued instruments, the classification may not be immediately available — new bond ISINs may be traded for several months before FIRDS data is populated.

Classification logic complexity: For bonds, the liquidity test under RTS 2 Article 13 and Annex III considers multiple factors simultaneously: asset class and sub-class (sovereign, covered, corporate, other), currency, average daily notional amount, average daily number of trades, and percentage of days traded. The test is not a simple threshold: an instrument must meet multiple criteria simultaneously to be classified as liquid.

Issuer vs. ISIN granularity: A highly liquid government bond issuer (e.g., France or Germany) issues many individual ISINs across different maturities. Most of the on-the-run (recently issued) bonds are liquid; older off-the-run bonds in the same series may be illiquid due to lower trading volumes as the market concentrates in the new issuance.

Priya built a FIRDS lookup module for Farida that:

1. Downloaded the full FIRDS dataset daily from ESMA's public API.
2. Parsed and indexed the dataset by ISIN, with a local SQLite database for fast lookup.
3. Applied the RTS 2 liquidity test logic to each instrument in Farida's trade universe.
4. Flagged ISINs for which FIRDS data was unavailable (new issuances) and defaulted them to "illiquid" with a 24-hour review queue for manual classification.
5. Maintained a daily changelog of classification changes — instruments moving from liquid to illiquid or vice versa — so that the reporting logic could be updated automatically.

The module revealed that Farida's existing spreadsheet had 47 misclassified ISINs out of 312 — a 15% error rate. Of these, 31 were instruments that FIRDS classified as illiquid but Farida had treated as liquid (meaning they had been published too quickly — not a material compliance issue but a data quality concern). Sixteen were instruments that FIRDS classified as liquid but Farida had treated as illiquid, meaning they had been published on a deferred basis when near-real-time publication was required. The latter represented an APA transparency breach for those instruments.

The Remediation Architecture

Rafael and Priya designed and implemented a four-component remediation architecture over twelve weeks:

Component 1: OTC Trade APA Trigger

The root cause of the reporting gap — the manual trigger for OTC bilateral bond trades — was eliminated by reconfiguring Farida's trade capture system to automatically generate an APA report event for every bond trade, regardless of execution method (venue-executed or OTC bilateral). The APA report event was routed to the firm's APA middleware, which then applied the FIRDS classification logic to determine the appropriate publication timeline.

The previously manual process was replaced by automated event generation. No trade could complete the settlement workflow without generating an APA report event. The APA event had the same mandatory status as the ARM event.

Component 2: FIRDS Integration Layer

Priya's FIRDS lookup module was integrated into the APA middleware as a real-time classification service. For each trade, the middleware queried the FIRDS service with the instrument ISIN and received the current liquidity classification, the applicable deferral regime, and the LIS threshold (where relevant). The classification was logged alongside the trade record for audit purposes.

For ISINs not found in FIRDS (newly issued instruments), the system defaulted to illiquid treatment with an alert to the operations team, who would conduct manual classification within 24 hours using the RTS 2 criteria.

Component 3: ARM/APA Reconciliation

A daily reconciliation job was implemented that extracted the list of ARM submissions from the prior trading day and compared it against the list of APA submissions for the same day. Any ARM trade without a corresponding APA submission (or a documented deferral reason) generated an alert in the compliance workflow system. The alert required resolution — either confirmation that the APA submission had been made and the reconciliation was a data matching error, or initiation of a late APA submission and a breach reporting assessment.

The reconciliation job ran at 08:00 each morning, covering the prior day's trades. Late APA submissions could be made proactively before the AMF's oversight processes identified the gap.

Component 4: Deferral Workflow Management

For illiquid bond trades where deferral was applied, the remediation built an active deferral management queue. Rather than batching all illiquid trades for end-of-day publication as the prior system had done, the new system calculated the specific deferral deadline for each trade (end of day, 48 hours, or 4 weeks depending on notional and instrument type) and scheduled the APA submission accordingly. An alert fired 30 minutes before each deferral deadline for manual confirmation that the submission had been queued.

The ESMA FIRDS Liquidity Test in Practice: A Worked Example

During the engagement, Priya prepared a worked example for Farida's operations team training session to illustrate the FIRDS classification logic for corporate bonds under RTS 2, Annex III.

Instrument: XS2345678901 — a EUR 500 million 5-year investment grade corporate bond issued by a French telecommunications company

FIRDS data (from ESMA quarterly update):

Result: The bond fails two of the three criteria. Under RTS 2's cumulative test (all criteria must be met for liquid classification), the bond is classified as illiquid.

Implications for Farida: - Post-trade APA publication for OTC trades in this bond: end of trading day (standard illiquid deferral) for most trade sizes; 48-hour deferral if notional exceeds EUR 50 million. - No continuous SI quote obligation applies if Farida is an SI in this instrument (illiquid non-equity SI obligations are less prescriptive). - Pre-trade transparency: Farida may use the Negotiated Transaction Waiver for bilateral trades if conditions are met.

Priya noted a practical complication: "This bond might trade three times on a Tuesday — which looks liquid intraday — and then not at all for six days. The FIRDS classification is based on a six-month historical average. So the classification is backward-looking. A bond can have a burst of liquidity after an issuer news event but remain classified as illiquid because the six-month average is still below the threshold. Firms need to understand that FIRDS classification is not a real-time liquidity gauge — it is a regulatory determination based on historical data."

The Remediation Outcome

Four months after the AMF supervisory letter, Farida submitted its remediation report. The report documented:

• Zero APA reporting gaps in the preceding eight weeks (as evidenced by daily reconciliation logs showing 100% ARM/APA match rate).
• Full FIRDS integration covering 100% of the firm's traded bond universe, with daily classification updates and a 24-hour new issuance review queue.
• Retrospective analysis of the prior 12 months' trading, identifying all missed APA submissions, with late submissions made where technically permissible and breach assessments documented where deferral deadlines had passed.
• Revised operations procedures and trained operations team.

The AMF closed its supervisory inquiry without escalating to formal enforcement. The private supervisory letter remained the only regulatory output.

Rafael submitted his engagement report with a final observation: "Farida's failure was not the result of bad faith or indifference to regulatory obligations. It was the result of treating bond post-trade transparency as a simpler problem than it is. The equity post-trade reporting world has years of automation, standardised ISIN coverage, and well-established APA workflows. The bond world has none of that. Bond transparency compliance requires active, dynamic data management — FIRDS refresh, deferral deadline tracking, OTC trade capture — that cannot be treated as a set-and-forget implementation."

Discussion Questions

1. The root cause of Farida's reporting gap included a confusion between ARM and APA obligations — a misunderstanding that both obligations could be satisfied by the same submission. How would you design an onboarding training programme for a new operations analyst that makes the ARM/APA distinction permanently memorable? What analogies or examples would you use?

2. The FIRDS classification system is backward-looking — it uses historical average daily trading data to determine liquidity. What are the regulatory trade-offs of this approach? Can you think of situations where a backward-looking classification would cause a compliance error even for a firm doing everything correctly?

3. Farida's reconciliation gap — the absence of a daily ARM/APA match — was described as "the most preventable failure." Design a simple reconciliation logic in pseudocode or Python that would catch an APA reporting gap within 24 hours. What data sources would the reconciliation query?

4. The FIRDS data showed that 98%+ of corporate bond ISINs are classified as illiquid, meaning that the majority of the EU bond universe benefits from deferred publication. What does this mean for the informational value of the EU bond post-trade transparency regime? If most trades are deferred, can the transparency regime achieve its stated goal of improving price formation in bond markets?

5. ESMA's data quality analysis — comparing APA publications against ARM transaction reports — was what triggered the AMF supervisory inquiry. What does this tell us about the regulatory surveillance strategy? How does the existence of two parallel reporting streams (APA and ARM) create a natural audit trail for regulators even without active surveillance of individual firms?