Designing and Building a Structured Asset Model Suite for a Global Professional Services Firm

Objective

A global professional services firm serving the life insurance sector engaged Graeme Group to design, build, implement and document a suite of structured asset models within an enterprise actuarial modeling environment. This was a build engagement rather than a validation: the objective was to create modeling capability that did not previously exist, to a standard that would stand up when the firm’s own clients relied on it.

Structured assets are difficult to model well. A collateralized mortgage obligation does not behave like a corporate bond. Cash flows are governed by a contractual waterfall that allocates principal and interest across tranches according to rules that change with the performance of the underlying collateral, and the timing of those cash flows depends on prepayment behaviour that is itself a function of the rate environment. Getting a defensible projection out of that requires the waterfall logic to be implemented faithfully, the collateral assumptions to be economically sensible, and the two to interact correctly under stress as well as at base.

Insurers hold these instruments in volume, and the capital and reserving frameworks that sit over them increasingly demand asset cash flows projected with the same rigour long applied to liabilities. Any actuarial platform intended for the life sector therefore has to model them credibly. The firm needed that capability built, benchmarked against independent sources, documented to a standard a third party could review, and handed over in a state its own team could maintain.

Scope of Work

The engagement covered four structured asset classes, collateralized mortgage obligations, mortgage-backed securities, asset-backed securities and collateralized loan obligations, taken from design through to implementation, benchmarking, documentation and post-deployment support.

The work was deliberately sequenced rather than run in parallel. A cashflow and waterfall framework was designed and built for the first asset class, benchmarked, and only then extended to the remaining three. Structured assets share more mechanics than their labels suggest, and building one properly first produces a framework the others can inherit; building four at once produces four one-off models that diverge on maintenance. The sequencing was a design decision about the shape of the finished suite, not a scheduling convenience.

Scope also expressly included the surrounding work that determines whether a model operates reliably in production: configuration of the platform’s asset modules, integration with the components and workflows that feed and consume the models, methodology documentation, and support through the firm’s own user acceptance testing and the period immediately after deployment.

Services Provided

Cashflow and Waterfall Design

The first phase built a cashflow and waterfall modeling framework from the contractual mechanics up: how principal and interest are allocated across tranches, how that allocation shifts as collateral performance moves, how credit enhancement and subordination absorb losses, and how prepayment assumptions drive the timing of everything downstream. The framework was designed to be parameterized rather than hard-coded, so that instrument-specific structures could be represented as configuration rather than as new code.

Implementation and Extension

The framework was implemented within the platform’s asset modules, including the calculation logic, the data and parameter setup, and integration with the surrounding workflows. It was then extended and adapted to cover the remaining three asset classes, with the explicit goal of a consistent and maintainable suite rather than four separate builds. Where an asset class required different treatment, that divergence was intentional and documented; where it did not, the shared framework was reused.

Benchmarking Against Independent Sources

Each implemented model was reconciled against the firm’s own reference calculations and against independent market analytics, comparing key cash flows, risk metrics and valuation outputs. This step establishes that the model is right rather than merely running, and it is the step most often compressed when a build runs late.

One boundary should be stated clearly. This was self-benchmarking of our own build: confirmation that what we implemented reproduces independent reference points. It is not independent model validation in the sense of ASOP 56 or a second-line model risk review, and we did not present it as such. A party that builds a model should not be the party that independently validates it, and the documentation was written so that a genuine independent validation could be performed later without re-deriving the methodology from scratch.

Documentation and Handover

Comprehensive model documentation was produced alongside the build rather than retrofitted after it: methodology, assumptions, inputs and outputs, configuration choices, integration points and controls. The test of this documentation was whether someone who had not built the models could maintain, extend and defend them. We also supported the firm through user acceptance testing and the initial post-deployment period, analyzing test results, responding to queries and making agreed refinements to the models, configuration and documentation.

Deliverables

• A parameterized cashflow and waterfall modeling framework covering collateralized mortgage obligations, extended to mortgage-backed securities, asset-backed securities and collateralized loan obligations as a single consistent suite.
• Full implementation within the client’s actuarial modeling environment, including calculation logic, data and parameter configuration, and integration with the surrounding components and workflows.
• Benchmark reconciliations of each implemented model against the client’s reference calculations and independent market analytics.
• Methodology and design documentation for each asset class, written to be maintainable and reviewable by parties who did not build the models.
• Support through client-led user acceptance testing and post-deployment hypercare, including issue analysis and agreed refinements.
• Knowledge transfer to the client’s own modeling team.

Outcome

• Capability That Did Not Previously Exist: Four structured asset classes moved from unmodeled to implemented, benchmarked and documented within a single engagement, delivered to a fixed budget across roughly fifteen weeks.
• A Single Consistent Suite: Because the framework was designed once and extended three times, the client received a consistent and maintainable set rather than four independent builds with four separate maintenance burdens.
• Documentation Written for Future Maintainers: The methodology documentation was written for a reader who was not present during the build, which is what makes a model an asset of the firm rather than an asset of the person who wrote it.
• A Clear Boundary on Assurance: The engagement included benchmarking of our own build, stated as such in the documentation. Independent validation remains a separate exercise, and the documentation was written to make that exercise efficient.
• Follow-On Scope Specified: The engagement closed with a fuller reconciliation program scoped and specified in detail, giving the client a costed decision rather than an unfinished workstream.

This engagement shows Graeme Group building actuarial modeling capability, not only reviewing it. Structured assets reward teams that understand both the contractual mechanics of a waterfall and the practical realities of implementing one inside a production modeling environment, and the same understanding that makes our validation work rigorous is what makes our build work maintainable.