A business banking ECL engine the risk team runs itself
A UK bank asked us to build an expected credit loss engine its business banking risk team could run itself.
Every business banking ECL run sat with Finance, so modellers could not test a change within a development cycle.
What We Provided
A known baseline, so every later change can be trusted
Rebuilding the existing SAS engine in Python and proving an account-level match gave every subsequent change a baseline to measure from, and gave the bank the confidence to move at all.
One file that defines a run, and can be reviewed
One configuration file assembles the models, passes the parameters and lookup tables and defines the outputs. A run is described by one reviewable artefact rather than reconstructed from a process.
Asking what a different economic view would cost
Models grew from lookup tables into term-structure calculations, so the team can feed in a different economic input file and see the effect. Risk committees ask that question often.
Usable by more people than the modeller who built it
Analysts drive a run from a notebook and render an HTML report of each stage, and a no-code interface surfaces the controls a monthly run requires.
The Results
The bank's own analysts run the engine on ordinary hardware, reproducing the SAS results account for account. One reviewable configuration file defines each run, and the HTML render lets a reviewer follow any number back to its source.
Other Case Studies
Independent validation of a Pillar 2A concentration add-on
A European bank needed to know that the capital it holds against concentrated lending would stand up to scrutiny, so it asked us for an independent view of the model behind the number.
Redeveloping every IFRS 9 model across two portfolios
A UK lender needed every model in its IFRS 9 expected credit loss suite rebuilt, at the point when the team that had built them was no longer there.