Scenario-Based Machine Learning (SBML) is emerging as a powerful refinement to traditional portfolio optimisation techniques used across the insurance industry. Instead of relying on simplified, linear assumptions, SBML applies machine learning to thousands of stochastic scenarios, enabling models to detect complex interactions, non-linear responses and multi-objective trade-offs. The approach gives insurers a dynamic framework capable of reflecting regulatory capital demands, liability structures and shifting risk appetites more realistically than conventional optimisation tools.
SBML is particularly valuable for insurers because they must juggle competing objectives—such as returns, solvency requirements and market risk constraints—while accounting for intricate balance-sheet interactions.
The methodology allows insurers to assess portfolios in a forward-looking way, stress-testing strategies under a wide range of economic pathways. This richer analytical lens provides a clearer understanding of how portfolios behave under real-world conditions, giving investment teams deeper insight into risks and opportunities.
Ortec Finance has teamed up with Insurance Asset Risk to examine how SBML can help insurers modernise their portfolio construction processes.
In an on-demand webinar, Ortec Finance senior business specialist Ashish Doshi and M&G Investments insurance solutions specialist Iain Ritchie explored how SBML uses stochastic scenarios and advanced learning techniques to capture intricate portfolio interactions. By modelling non-linear behaviour, SBML provides insurers with a more accurate and dynamic understanding of how portfolios may perform under different market regimes.
Case study phase one: building the efficient frontier
The initial joint case study with M&G Investments focused on maximising surplus mean and the 5% Conditional Value at Risk (CVar), while keeping the market risk SCR charge stable.
“SBML produced an efficient frontier of viable portfolios, whereas traditional optimization often yielded only a single usable solution,” Ortec Finance senior business specialist Ashish Doshi said.
According to Ritchie, “One of the main advantages of SBML modelling is the efficiency it brings. It allowed us to identify viable portfolios quickly and focus on strategic client discussions.”
Case study phase two: broadening optimisation goals
The second phase introduced a third objective—optimising mean, 5% CVar and market risk SCR simultaneously—creating an efficient plane rather than a single frontier.
“The ability to optimise across returns, risk, and capital will be a significant enhancement,” Ritchie said. “It helps us assess trade-offs more holistically and identify portfolios that strike the right balance.”
Aligning advanced modelling with human expertise
Both speakers emphasised that SBML enhances judgement rather than replacing it. Contribution analysis, explainability tools and stress testing remain central to ensuring model outputs are transparent and decision-ready. SBML ultimately strengthens insurer decision-making by pairing machine-driven insight with human oversight.
Copyright © 2025 FinTech Global
