PD models
Through-the-cycle and point-in-time estimation, term structures, and calibration appropriate to portfolios with limited default history, which is most portfolios in this region.
Advisory
Expected credit loss models that survive the audit.
We build, calibrate and validate IFRS 9 models for banks, microfinance institutions, credit providers and insurers — and we have done it forty-nine times across seven markets.
Context
What you are facing
IFRS 9 is not one model. It is a chain of judgements, each defensible on its own and collectively hard to explain under pressure.
Where exactly does significant increase in credit risk begin, and can you show why? Is the thirty-day backstop a trigger you rely on or one you have rebutted, and on what evidence? How does a through-the-cycle PD become point-in-time, and does the term structure hold at year three? What recovery assumption sits inside your LGD, and when was it last tested against actual recoveries? How are the macroeconomic scenarios weighted, and would you defend those weights to a supervisor who thought them optimistic?
Most institutions can answer these. Fewer can evidence the answers quickly, in a form an auditor accepts, using a model that has not drifted since the year it was built.
Deliverables
What we deliver
Concrete outputs, not activities.
LGD models
Recovery curves built from your own workout data, collateral haircuts, time-to-recovery discounting, and treatment of restructured exposures.
EAD models
Including credit conversion factors for undrawn commitments and revolving facilities.
Staging and SICR methodology
Quantitative and qualitative triggers, backstop treatment, cure periods, and the documented rationale for each threshold.
Macroeconomic overlay
Scenario definition, the statistical link between macro variables and observed default behaviour, and probability weighting you can defend.
Model documentation
The pack your auditor, your validator and your successor all need, written to be read rather than filed.
Independent model validation
Of models built by other vendors or in-house. We do not validate our own.
Post-model adjustments and their governance
Because every institution has them and few can explain how they are approved, sized or released.
Methodology
How we work
Step 01 of 04
01
Discover
We start in your loan book. Data quality determines what is modellable, and finding that out in week two is considerably cheaper than finding it out in month four.
Jurisdiction
We work to the requirements of the Central Bank of Kenya and SASRA, the Bank of Uganda, the Bank of Tanzania, the National Bank of Rwanda, the Bank of Botswana and the Bank of Zambia — and to the prudential guidance each issues on impairment and provisioning.
Where we have done this
Across a national microfinance sector
Working through a microfinance association, we delivered IFRS 9 to fourteen of its member institutions — a sector-wide implementation rather than a series of separate engagements.
At a central bank
IFRS 9 implementation for a regional central bank, alongside a data analytics programme delivered to its CFO forum.
Beyond banking
Insurers, a reinsurer, a pension services group, a credit reference bureau, a SACCO and a wholesale lender — institutions holding financial assets under IFRS 9 whose portfolios look nothing like a bank’s.
Related services
The deepen motion — where existing clients extend.
Next step
Tell us where the challenge is coming from.
An auditor’s question, a supervisory finding, a model that has not been touched in three years, or a first implementation. Describe it and we will come back with an approach, a timeline and a cost.