Pricing — the data model

Start with one node. Open the sample price program. It is a priced Program with the id PROG1. Read its content top to bottom and you have read the whole module:

header            Reinsurance/Samples/PropertyCatTreaty            ← the treaty it prices
acceptancesById   PropertyCat2023  → Reinsurance/Acceptance/PropertyCat2023
                  PropertyRisk2023 → Reinsurance/Acceptance/PropertyRisk2023
sectionsById      S1               → Reinsurance/Samples/PropertyCatSection
lossModelsById    property-cat-2024 → LossModelling/FrequencySeverityLossModel
simulationSettings  10,000 simulations, compressed to 5, last run 2024-03-01 09:30, 4,200, "completed"
simulationResults   S1-GrossLoss  [20, 45, 70, 95, 140]
                    S1-CededLoss  [ 0,  0, 10, 35,  40]

Every value on the right is a node path: into the Reinsurance partition for the treaty, acceptances and sections, and into Loss Modelling for the loss model. The Program never holds a C# reference to a Reinsurance or Loss Modelling type. That is why the dependency is strictly one-way: Pricing → Reinsurance + LossModelling.

The two result distributions are illustrative draws, related by the section's 40 xs 60 layer alone. Each ceded draw is min(max(gross − 60, 0), 40). The section's annual aggregate deductible (5) and limit (120) are not applied, so these are pre-aggregate figures:

The statistics over them (mean, standard deviation, quantile) are on the Formulas page.

The picture, drawn by the platform

Prerequisite: the class diagrams on this page are embedded from the Data Modelling plugin's explorer (/DataModelling/Explorer), which Pricing does not require. On an installation without it they render nothing; each type's own page under /Pricing still lists its properties.

The entities

Program: the pricing root

Source: Pricing/Program/Source/Program.cs. Type node: /Pricing/Program.

The opaque-id bridge. The sample section PropertyCatSection carries models: [{ lossModel: "property-cat-2024", sections: [...] }]. That is a LossModelMap (Reinsurance/Source/LossModelMap.cs). It knows the id only, never which model it is. The Program's LossModelsById records which model the id resolves to. Rebinding it changes one Program entry and leaves the contract untouched. It is not a re-price: Pricing ships no runner that follows this path and recomputes the distributions. The stored SimulationResults stay exactly as they were until something writes new ones. In the sample the id resolves to the type node LossModelling/FrequencySeverityLossModel, because no frequency-severity instance node ships (see the Loss Modelling data model).

The referenced nodes:

• the treaty Alpina Property Cat XL 2024 (TR-2024-001)
• the acceptances Property Cat 40 xs 60 — UWY 2023 and Property Risk 20 xs 10 — UWY 2023
• the section Property Cat XL (S1): 40 xs 60, AAD 5, AAL 120

For the contract side, see the Reinsurance data model.

SimulationSettings: the run configuration

Source: Pricing/SimulationSettings/Source/SimulationSettings.cs. It is embedded in the Program.

SimulationResult: one simulated distribution

Source: Pricing/SimulationResult/Source/SimulationResult.cs. It is embedded in the Program.

The key of a result is (StructureId, AmountType, Period) plus its Id. StructureId joins it to SectionsById.

Guideline: the underwriting guideline as rules

Type node: /Pricing/Guideline. Its content type is PricingState (Pricing/Guideline/Source/PricingState.cs), a priced placement. The rules themselves are the Guideline scope (Pricing/Guideline/Source/Guideline.cs) evaluated over that state.

The outputs are not stored. They are: Escalation (a flags enum: SecondUnderwriter, HeadOfLoB, Cuo, UnderwritingCommittee, Legal), GuidelineVerdict (Write, ReferToCuo, Decline), GuidelineShareAction (Maintain, Increase, Decrease, Decline), and one RuleResult per clause (Clause, Title, Status, Messages, Escalation). See the Formulas page.

How it hangs together

Nothing in Reinsurance or Loss Modelling points back at Pricing. The test ProgramTests → "no back-reference into Pricing" pins that.

Check yourself

Q. An underwriter wants PROG1 to use a different loss model. The new model sits at LossModelling/FrequencySeverityLossModel/PropertyCat2025. Which node changes, and which does not?

A. Only the Program changes: lossModelsById["property-cat-2024"] gets the new path. The section PropertyCatSection keeps saying lossModel: "property-cat-2024". It holds the opaque id, never the path, so the contract is untouched.

Rebinding is not the re-price itself. Pricing has no runner that resolves the path and recomputes the distributions, so the stored simulationResults stay unchanged until a run writes new ones.

Reference

Next: the formulas · Loss Modelling data model · one distribution ceded through QS, XL, tower and stop-loss.