Credit Risk Analytics

At GPS.ai, we are using machine learning and artificial intelligence to transform the banking industry, using vast amounts of data to build models that improve decision making, tailor services, and improve risk management.

We help clients optimize their risk exposures, improve performance, increase profits, and accelerate growth.

Our advanced credit risk analytics enable institutions to improve underwriting decisions and increase revenues while reducing risk costs. We work across all asset classes, credit risk models, and the entire credit life cycle, including profit maximization, portfolio management, and loss mitigation.

Stress Testing

Our stress-testing capabilities include generating scenarios, translating them into environmental parameters through macroeconomic quantification, and then assessing the impact of these scenarios on the market, and on the client’s profit and loss (P&L) and balance sheet. All of this informs an action plan to mitigate risks and swiftly capture opportunities.

We help clients complement traditional risk analytics with machine learning to find previously unidentified patterns and make better predictions.

Risk Modeling

Most BHCs have had difficulty in producing CCAR PPNR ICAAP models that have an appropriate level of statistical rigor. Without statistical rigor the models will not be accepted by the Fed during stress testing.

Getting better at CCAR PPNR ICAAP models requires making a realistic assessment about where a BHC is today and where it needs to be in several years, and then mapping out a game plan to get there.

BHCs are rightfully proud of their ability to understand and model credit losses. For most, modeling the entire balance sheet and income statement using an analytical and statistically rigorous approach is still relatively new, and it will require a concerted program to bring CCAR PPNR ICAAP modeling up to where regulators believe it needs to be.

Our experts follow the following key considerations during CCAR PPNR ICAAP modeling.

1. Segmentation: How do we ensure optimal segmentation to balance the conceptual soundness of segments but avoid going to overly granular levels where noise in data leads to model robustness issues?

2. Estimation Metric: What is the optimal estimation metric e.g., end-of-period balance vs cash-flow components of balances, revenues vs underlying activity business driver?

3. Historical Data: What are the minimum historical data requirements required to develop robust stress testing models? What are the potential solutions for banks with historical data limitations?

4. Model Structure: What is the appropriate model structure considering the trade-off between statistical accuracy and conceptual soundness?

5. Statistical robustness: What are the expectations for statistical robustness for stress testing models and potential tests to demonstrate it?

6. Output analysis: How do we ensure that the outputs reflect an informed view of the business and are consistent with actual results based on historical data?

7. Sensitivity analysis: How can sensitivity analysis be used to test model robustness and identify potential mitigations for model limitations?

8. Documentation: What are the expectations for documentation to ensure that the process is transparent and repeatable?

CCAR / DFAST / STRESS TESTING

CCAR / DFAST / Stress tests are cost-intensive endeavors. They demand many dedicated employees and the active involvement across virtually the entire organization – risk, finance, treasury, business units, audit, senior management, and the board of directors. Additionally, a failure to comply with a regulator’s requirements could result in severe penalties, for example, a mandatory recapitalization, forced sale of operating units and portfolios, and management changes. It has therefore become a key priority for banks to develop an effective and cost-efficient approach to stress testing.

We use our experience and expertise to help banks develop better stress tests.

Our 6-step process will help you achieve success in your stress testing endeavors.

Step 1: Data Preparation

• Standard and clear data structure

• Validated data integrity (reconciled with balance sheet)

• Dedicated data team

Step 2: Strategic PMO

• Single point of contact with evaluator

• First-time-right response to questions

• Adequate senior leadership and attention

Step 3: Robust Methodologies

• “Certified” impairment & Estimated Loss (EL) methodology& Loss Estimation strategies

• Thorough forward-looking P&L and balance sheet forecasting using time series analysis

• Forward-looking regulatory-ratio-projection methodology

Step 4: Superior Portfolio Analytics

• Granular bottom-up portfolio stress-testing models

• Capabilities to reverse engineer key assumptions

• Deep understanding of (asset) valuation and classification

Step 5: Macro-scenarios and correlations

• Strong in-house capabilities to develop economic scenarios

• Econometric models linking macro-economic parameters to portfolio risk factors and bank-performance drivers

Step 6: Stress-Testing Model

• A stress testing model aggregating individual portfolios

• Asset-liability modeling functionality

• Advanced capabilities to run “reverse stress tests”

• Concrete management action plan coupled with strong board governance