Precise model performance through continuous monitoring

FRTB Model Monitoring & Re-Calibration

Continuous monitoring and re-calibration of FRTB risk models is essential for sustainable compliance and capital efficiency.

  • 01Continuous validation of model performance
  • 02Proactive re-calibration in response to market changes
  • 03Optimization of capital requirements through precise models
  • 04Automated backtesting and monitoring processes
11+Years of experience
120+Employees
540+Projects
ISO 27001certified

Model Monitoring as the Foundation of FRTB Compliance

Implementing FRTB significantly raises the bar for the ongoing quality assurance of internal market risk models: daily backtesting, P&L attribution at trading desk level and monitoring the modellability of risk factors determine whether desks retain their Internal Models Approach approval. When model performance deteriorates, banks face capital surcharges or a fallback to the standardised approach - with direct consequences for capital requirements.

Our service covers the setup of automated monitoring infrastructures, statistical validation of model quality, data-driven re-calibration strategies and the associated governance and documentation. The result: you detect model deviations early, re-calibrate in a targeted way and sustainably protect your IMA approval and capital efficiency.

2 service modules

Service scope

01

Continuous Model Monitoring

Systematic monitoring of the performance of your FRTB risk models through automated processes and statistical validation procedures.

  • Daily backtesting routines for Expected Shortfall models
  • Implementation of traffic light systems for model performance
  • Development of Model Performance Indicators (MPIs)
  • Automated alerting systems for model deviations
02

Strategic Re-Calibration

Proactive and data-driven re-calibration of your risk models to optimize forecast quality and capital efficiency.

  • Market regime-specific calibration strategies
  • Optimization of lookback periods and decay factors
  • Integration of forward-looking elements into model calibration
  • Validation and testing of new model parameters

5 phases

Our Approach

We work with you to develop a systematic approach to model monitoring that meets regulatory requirements while maximizing operational efficiency.

  1. Establishing solid monitoring infrastructures with automated alerts

  2. Implementing statistical tests for early detection of model deviations

  3. Developing data-driven re-calibration strategies

  4. Integrating stress testing into continuous model validation

  5. Building comprehensive documentation and governance processes

Your contact

Melanie Düring

Head of Risk Management

Continuous monitoring and re-calibration of FRTB models is a critical success factor for sustainable compliance. With a systematic approach, banks can not only ensure regulatory certainty but also optimize their capital efficiency through precise risk models.
ADVISORI expert opinion

6 QUESTIONS, BRIEFLY ANSWERED

Frequently asked questions about FRTB Model Monitoring & Re-Calibration

Why do FRTB models need regular re-calibration?

Market conditions change continuously

• volatility regimes shift, correlations break down in stress phases, and new products alter portfolio composition. Expected Shortfall models are calibrated to historical observation periods, so their parameters inevitably drift away from current market reality over time. In addition, the modellability of risk factors can change as the availability of real price observations evolves, moving factors between modellable and non-modellable status with direct capital impact. If this drift is not addressed, it surfaces as backtesting exceptions and deteriorating P&L attribution results, which in turn trigger higher capital multipliers or the loss of Internal Models Approach eligibility at desk level. Regular, data-driven re-calibration keeps forecast quality high, capital requirements efficient and model approval secure.

How often should FRTB models be reviewed and monitored?

FRTB effectively imposes several overlapping review cycles rather than a single annual check.

🔍 Established monitoring rhythm:

• Daily: backtesting of VaR forecasts against actual and hypothetical P&L at bank-wide and desk level
• Quarterly: evaluation of P&L attribution test metrics for each trading desk in scope of the internal model
• Annually: comprehensive independent model validation covering methodology, data and implementation
• Ad hoc: reviews after significant market stress, methodology changes or the approval of new products

In practice, the decisive factor is not the formal cycle but how quickly threshold breaches become visible. We therefore recommend automated monitoring dashboards with alerting, so deviations are escalated the day they occur rather than discovered at quarter-end.

What role does backtesting play under FRTB?

Backtesting is the central quantitative control for internal model quality under FRTB. Banks compare their one-day Value-at-Risk forecasts at the 97.5% and 99% confidence levels against both actual and hypothetical P&L

• and, unlike under the previous market risk regime, this happens not only bank-wide but at individual trading desk level. Exceptions are counted over a rolling twelve-month window. If a desk exceeds the defined exception thresholds, it loses its eligibility for the Internal Models Approach and must calculate capital under the standardised approach, which is typically significantly more expensive. At aggregate level, accumulating exceptions increase the capital multiplier. A robust backtesting infrastructure with clean P&L data is therefore a direct lever for capital efficiency, not just a regulatory formality.

What happens if a trading desk fails the P&L attribution test?

The P&L attribution test (PLAT) compares the risk-theoretical P&L produced by the risk model with the hypothetical P&L from front office systems, using two statistical metrics

• the Spearman correlation and the Kolmogorov-Smirnov test. Based on the results, each desk is assigned to a green, amber or red zone. Desks in the amber zone remain in the internal model but incur a capital surcharge; desks in the red zone must switch to the standardised approach until their test results recover. In practice, PLAT failures are rarely caused by the model itself
• most stem from data and valuation misalignments between front office and risk systems. A structured root-cause analysis of data lineage, pricing models and risk factor coverage is therefore the first remediation step.

What data do we need for effective FRTB model monitoring?

Effective monitoring stands or falls with data quality and consistency across systems.

🔍 Core data requirements:

• Daily P&L series per desk: actual, hypothetical and risk-theoretical P&L on a consistent valuation basis
• Complete risk factor time series with sufficient history, including stressed periods for Expected Shortfall calibration
• Real price observations to assess risk factor modellability and identify non-modellable risk factors
• Position and sensitivity data reconciled between front office and risk systems
• Model metadata: parameter versions, calibration dates and change history for governance and audit purposes

Beyond availability, lineage matters: supervisors expect banks to demonstrate where monitoring data comes from and how it is quality-assured. We help you build these data pipelines and controls as part of the monitoring infrastructure.

Can FRTB model monitoring and re-calibration be automated?

To a large extent, yes

• and automation is the only realistic way to handle the daily monitoring frequency FRTB demands at reasonable cost. Statistical tests, backtesting engines, threshold-based alerts, modellability assessments and standard reporting can and should run automatically. What remains a human task is the interpretation: root-cause analysis of exceptions, the decision whether an observed deviation warrants re-calibration or a methodology change, and the governance sign-off required for model changes. Our approach combines both layers
• we build automated monitoring pipelines integrated with your existing risk architecture and establish clear escalation and decision processes on top, including the documentation that model risk management and supervisors expect. This reduces manual effort while keeping model ownership and accountability where they belong.