RisCook reads a model's type, purpose, and inputs, then returns the specific statistical and ML challenger methodologies your validation team should run — with the rationale a regulator will ask for.
No black box. Every recommendation traces back to a documented reason your audit trail can show.
Tell RisCook the model's type, business purpose, inputs, and the segment it scores — or pull it straight from your model inventory.
RisCook returns the statistical and ML challenger methods appropriate for that model class, ranked by relevance and regulatory expectation.
Each test ships with the reasoning behind it, so your write-up explains why a test was chosen, not just that it was run.
Each archetype carries its own validated set of challenger methodologies, built from regulatory guidance and validation literature.
Scorecards, logistic regression, survival models.
Historical simulation, Monte Carlo, parametric.
XGBoost, LightGBM, shallow neural nets.
Staging models, macro overlay sensitivity.
Isolation forests, autoencoders, rules hybrids.
Black-Scholes variants, lattice, simulation.
Every recommended test links to the guidance or literature that justifies it — ready to drop into your validation report.
Every recipe generated is versioned, timestamped, and tied to the model card that produced it.
RisCook recommends. Your validators decide. Override any test, and the override itself is logged with a reason.
RisCook is currently onboarding model risk teams at banks, insurers, and asset managers.