Jabatix provides a performance database that stores information about financial assets in default as well as financial assets that have not defaulted.

The source of recovery is documented in the case of recoveries. This enables appropriate consideration of collateral during the calculation of risk provisions.


The history is supplemented by micro- and macroeconomic boundary parameters, which are often drivers of default. Customer and contract data are placed in an overall economic context.


The recovery database can be used as underlying for statistical methods for PD and LGD calculation. It is part of the result layer which already contains various data marts for specific analysing purposes.

The performance database created in this way, is then used as the basis for classic approaches as well as machine learning.


The source for the performance database are cash flows that are part of the contractual view as well as real payments (Core Cash Flow Generation).


For stage 3 impaired deals, the recovery database contains the expected recovery cash flows as well as real payments.

This data mart shows how much recovery compared to the expected recovery has actually been achieved.

The recovery database contains following data:

  • Information on all direct costs incurred after the moment of default on observed account/client (all collection costs - letters, SMS, FTE collection costs etc.; lawyer costs, court fees, ...)
  • All payments/cash flows incurred, starting from the moment of entering default - must be allocated to the source - debtor, co-debtor, collateral ID.
  • All information on collateral which are needed for SRP calculation - market values, type, location, size, ID, info on all re-evaluations and dates, internal haircuts, etc.
  • Effective interest rate in the moment of entering default as all costs and recoveries after default are discounted using that rate at the moment of default.
  • All mentioned inputs are available for the history and are connected to the exact Customer ID and Account ID together with the date when any event occurred (payment, costs, reevaluation, change of haircut).

If the real payments, which are automatically delivered to the solution, do not reflect the actual situation, users have the option of manually capturing real payments which will be uploaded into the data mart. A list is available for real payments captured manually in the past. It is possible to delete records in the list. Such "corrections" will be considered again from the next run of the data mart.


  • No labels