Banking & Market Risk8 min read

FRTB Implementation: Lessons for Trading Desk Risk

By Jonas Adam Mohamed Osman, known as Yonas Osman

CEO of Quantica Risk

60 Rue François Ier, 75008 Paris, France

Published 15 October 2026

What banks are learning as FRTB internal model and standardised approach deadlines approach, and how desks are restructuring to pass tests.

Executive answer

The Fundamental Review of the Trading Book replaces value-at-risk with expected shortfall calibrated to periods of market stress, and it forces banks to prove, desk by desk, that their risk models actually explain daily P&L before those desks can use internal models for capital. Many banks are discovering during parallel running that a meaningful share of trading desks fail the P&L attribution test, meaning they default to the more punitive standardised approach regardless of model sophistication.

This has turned FRTB from a quantitative modelling exercise into an operational and data-architecture challenge: risk factor eligibility, trade booking consistency, and the completeness of historical pricing data now determine capital outcomes as much as the statistical model itself. Banks furthest along in implementation are those that tackled data lineage and desk structure before refining the models themselves.

The P&L attribution test as the real bottleneck

FRTB's P&L attribution test compares a desk's risk-theoretical P&L, generated by the front-office pricing model feeding the risk engine, against the hypothetical P&L used for backtesting. Desks that fail on a rolling basis are pushed to the standardised approach, which the Basel Committee designed deliberately to be more conservative, creating a strong incentive for banks to fix underlying data and model consistency issues rather than accept the capital penalty. In practice, many failures trace back to risk factors that are mapped or approximated differently between front-office and risk systems, not to flawed statistical risk modelling.

Remediation typically involves reconciling the risk factors used in pricing against those used in the risk engine, tightening the universe of eligible risk factors through the non-modellable risk factor test, and in some cases restructuring trading desks around lines that can be modelled coherently. Banks that treated this as purely a risk-model problem rather than a front-to-risk data integration problem have generally seen slower remediation progress.

Standardised approach as a strategic fallback, not a failure state

The revised standardised approach, built on sensitivities-based risk weights across delta, vega, and curvature, is more risk-sensitive than its predecessor and, for some simpler or lower-volume desks, is now a deliberate strategic choice rather than a fallback forced by model failure. Running the standardised approach avoids the ongoing cost of internal model maintenance, approval, and attribution testing, and some banks are consciously limiting internal model scope to the desks where the capital benefit clearly outweighs that operational burden.

This reframing matters for resource allocation: rather than pursuing internal model approval across every desk, banks are prioritising the trading lines with the largest capital gap between standardised and modelled outcomes, and accepting standardised treatment elsewhere. Supervisors, including under the EBA's regulatory technical standards, have generally been receptive to this selective approach provided the desk structure and risk factor mapping remain transparent and well documented.

Non-modellable risk factors and data sufficiency

A recurring implementation challenge is the non-modellable risk factor test, which requires a minimum number of real, verifiable price observations for a risk factor to qualify for the internal model approach; factors that fail are capitalised through a punitive stress scenario charge. Less liquid instruments, emerging market curves, and bespoke structured products frequently fail this test, and banks are investing in third-party pricing data consortia to bolster observation counts where internal transaction data is too thin.

Getting this right requires close coordination between market data teams, model validation, and the trading desks themselves, since the classification of a risk factor as modellable or not can materially swing desk-level capital. Banks that built shared, well-governed market data utilities ahead of FRTB go-live have had a visibly easier path through supervisory review than those treating data sufficiency as an afterthought to model build.

Conclusion

FRTB implementation confirms a broader pattern in post-crisis regulation: the hardest part is rarely the statistical model itself but the consistency of data and processes feeding it. Banks that invested early in front-to-risk reconciliation and desk-level governance are capturing internal model benefits where they matter, while using the standardised approach pragmatically elsewhere rather than chasing full internal model coverage at any cost.

Jonas Adam Mohamed Osman, known as Yonas Osman, CEO of Quantica Risk, encourages trading risk teams to treat FRTB readiness as a data governance programme with a capital model attached, not the reverse, which tends to shorten the path to sustainable internal model approval.

Primary sources

FRTBmarket risktrading bookP&L attributionstandardised approach