Modules/Insight Core

Quant ALM Insight Core

The bank's models, made dynamic: calibration of the behavioural models that feed the cash-flow engine, variabilisation for simulation, documentation and versioning.

Role, inputs, outputs

Your behavioural models, industrialised and versioned.

Inputs
Historical dataQuantitative models
Quant ALM Insight Corequantitative modelling
Outputs
Calibrated, versioned modelsTraced variantsAutomated scatter-plot analysis
Role

Behavioural modelling: calibration of the quantitative models (stochastic and statistical) that feed the cash-flow engine and the other modules: non-maturity deposits, regulated savings, prepayments, customer options…

Inputs & outputs

Inputs: historical data, quantitative models. Outputs: calibrated run-off models and behavioural parameters, documented and versioned, subject to validation by the bank's teams.

Dynamisation & variabilisation

Model variants on demand, compared with the validated model, impact measured across the whole chain.

Principle

The models remain the bank's: the platform industrialises calibration, variabilisation and documentation; it does not replace internal governance with its own method.
A model library is also available.

ICAAP · ILAAP

Contribution to the ICAAP and ILAAP

Consistent cash flows and models for both assessments.

Functional scope

ALM and risk.

ALM and risk modelsModel variants on demandVersioning and documentationComparison to the validated modelModel governance
In conversation

Calibrate a model, read the fit.

A prepayment model can be calibrated straight from the chat. RIALTO fits the model on the dataset, compares the modelled rate with the observed rate and publishes the model. The run-off module then applies it to each contract, with its client rate and the reference rate.

rialto-kernel · chat Quant ALM
Calibrate a prepayment model on this dataset
RIALTO · Kernel
Calibration complete. R²: 0.71 · model published for the run-off as “Prepayment — fixed-rate mortgage”.
Observed vs modelled rate167 periods
ModelObserved
1,60,80
200920122015201820212023
Illustrative example · fictitious data
Versioned
documented, comparable models
Your models
governance stays internal
Your data
calibration on your histories

See RIALTO reason over your numbers.

We simulate a full analysis with your data types and the scenario of your choice.