Service
Beyond the traditional gap analysis
In addition to generating high level directional estimates, the QIA is complimentary to the Capabilities Review service. The standard gap analysis output is an inventory of the identified gaps – including a qualitative assessment of each gap. The qualitative assessment typically uses standard metrics to “rank” the gaps. This qualitative evaluation can be used by management to prioritize areas of focus for enhancement.
Through execution of the QIA process, companies can get firsthand experience executing a lifetime credit loss model before the mandatory CECL adoption date. Where gaps have been identified, assumptions are made and documented. Multiple iterations can be run with different scenarios and assumptions for the identified gaps. The QIA allows the user to see the impact of the gaps in real time; the results serve to help identify which gaps are likely to have the greatest impact and priority for enhancement.
The QIA is customized based on the client’s interpretation of CECL and can be performed for the entire portfolio or on selected portfolios. The results are based on client data and the methods, assumptions and reasonable and supportable forecasts chosen by the client.
The CECL QIA leverages a standard three-step methodology:
1. Extract, transform, and load CECL input data from client systems
- Assess current data for both calculations and end-state reporting requirements
- Assess the legacy data feeds and source systems that will feed the credit risk models
- Identify gaps in current available data
2. Simulation of lifetime expected credit losses
- Obtain preliminary range of quantitative impacts of adopting a lifetime credit loss model
3. Analyze simulated lifetime expected credit loss results
- Determine the reasonableness of results
- Evaluate the key drivers related to the QIA results
- Evaluate the potential impact of different CECL implementation options
- Derive recommendations for managing the impact of transitioning to a lifetime expected credit loss model
