Finalyse risk advisory for insurance offers a comprehensive range of actuarial and risk management services to provide solutions tailored to unique challenges faced by insurers.

As an independent consultancy, Finalyse is agnostic in terms of modelling tools, and our aim is to support firms identify the best tool based on their requirements.

This blog post is intended to highlight the key features of one of the modelling tools, Insurance risk suite (formerly, Prophet), a proprietary tool offered by FIS^[1], to help insurers understand the importance of the skillset required to apply on such modelling tools.

Overview

Many insurance firms use proprietary modelling software to perform actuarial modelling. The speed and flexibility offered by FIS ® Prophet^[2] makes it a popular choice.

In this blog post, we aim to provide a high-level overview about the key and advanced features and applications of Prophet along with their practical user experience and challenges.

Furthermore, we discuss how the need to merge models has arisen over time as a result of new regulations and standards. We also discuss the key considerations when migrating to Prophet including reviewing and validating actuarial models and importantly, model governance considerations for Prophet models.

Key Features

Some of the features and applications offered with the Prophet workspace are highlighted on the chart below and we discuss the most popular ones amongst developers and users in more details subsequently:

Data Conversion System (DCS)

Data Conversion System (DCS) is an application widely used by Prophet developers that helps structure the raw data and produce model point (or data input) files in a format acceptable for Prophet.

It allows reference tables along with multiple input data files to be used to adjust and enrich data. The program is capable of producing single or multiple model point files for each product or product class and can be configured to output either individual policy level model point files for Prophet calculation runs or grouped policy model point files for Prophet grouping runs.

Diagram View

Diagram View is an important feature for a Prophet model developer as it makes it convenient to trace dependents and precedents of variable(s) in each library. It provides transparency that helps with debugging, troubleshooting and to gain deeper understanding of cashflow components and results. The diagram view also shows resulting values of variables when a completely executed run is selected for a product or an accumulation^[3].

Model point files, tables, modules, and variable properties can be inspected through the diagram view in the variable flow as shown on the chart below:

Filters can be applied in the diagram view as desired and convenient for modelling. For example, filter by dependent or precedent, filter elements of array dimension or exclude self-referencing variables, etc.

Schedule

Schedule is a useful feature that enables Prophet developers to form a view on key aspects of products modelled in the workspace. It is used when merging multiple Prophet models which were developed for various calculations like IFRS17 and Solvency II or where a business combination exists, to optimise resources and to streamline the library structure.

This produces a grid of variables (all or specific type) and products indicating the variable type, indicator or dimension, etc.

Parallel Processing

Parallel Processing provides additional benefit to attain higher performance than a single processor-based system. This means that multiple machines are connected to a network with a Master PC assigned and used to operate from. This enables leveraging the computing power of all the PCs appropriately connected via a server which results in reduced runtime by splitting either the products, simulations (e.g., in case of stochastic calculations) or multiple model runs across different processors.

Features like multi-threading are also useful to improve efficiency which comes along with the Enterprise solution (offered by FIS).

Advanced Features

Advanced features within the Prophet allow for complex calculations in an efficient manner. Some of these are discussed below:

The Asset Liability Strategy (ALS)

The Asset Liability Strategy (ALS) library supports investment decision making. The library provides a mechanism to model assets and liabilities together including complex interactions such as policyholder behaviour (e.g., dynamic lapses) and management actions (such as discretionary dividends, bonuses, or participation rates). It makes the asset liability models more dynamic while:

Measuring the impact of different investment strategies that include the re-balancing of asset portfolios
Measuring the value of new business and the impact of acquired portfolios
Calculating the risk-based capital, including Solvency II and Internal Model
Performing stochastic calculations to appropriately estimate the time value of options and guarantees, etc.

The ALS library is considered one of the most complex libraries, as it involves extensive code specifically defined in extended formula.

Extended Formula

Extended Formula are variable types, and these are either already defined in a library (e.g., the ALS library) or can be created to enable specific and more complex calculations than those carried out by the standard Prophet formula.

Examples for use cases of extended formula include the rebasing of guaranteed rates, the isolated calculation of each group of paid-up contracts and the increments in contract benefits.

Extended formula can be set up either as time dependent or non-time dependent given the requirements. It allows additional variables to be defined within the formula when they are not defined elsewhere in the Prophet model. Parameterisation within the extended formula can be used to optimise the model performance.

However, the variables defined within the extended formula are not displayed in the diagram view with associated dependencies and the values stored in the extended formula’s internal variables are not visible directly in the model, therefore debugging gets challenging at times.

Stochastic Calculations

Stochastic calculations involving multiple simulations can be performed using Prophet. They can be tailored to optimise shareholder and policyholder investment returns by understanding the interaction between bonus rates and future investment performance, to allow for guaranteed annuity rates or to perform insolvency and stress tests to meet the requirements of various capital regimes such as embedded value or Solvency II.

Setting up a stochastic run is different to a deterministic run. For example, it requires stochastic tables where assumptions vary by simulations with appropriate configurations for the run structure and settings.

Model Merge

There have been a lot of actuarial model developments recently due to the implementation of IFRS17 from 1 January 2023 and this was preceded by another wave of developments in Europe around the implementation of Solvency II on 1 January 2016. This has created decentralised models over the period within many organisations.

Maintaining and developing multiple models give rise to inconsistency in input and output, often leading to inefficiency due to duplication of effort, longer runtime, even misinterpretation of results. Maintaining an inefficient modelling architecture can become more challenging, especially if modelling resources are constrained.

One way to address the challenges mentioned above is to consolidate or merge models in order to:

Standardise pricing, business reporting, business planning and asset liability management processes and enable reporting on more than one basis using a single valuation model
Avoid the complexity of model dependencies which are holding back key modelling decisions
Easily implement model updates and include essential components that are required in models for all activities

Some of the key aspects to keep in mind while considering activities related to merging Prophet models include:

Use of comparison logs of libraries across different models
Use schedule to understand indicators and variable definitions while combining them
Import variables to the base model
Adjust indicators and create new indicators and/or variable definitions
Perform unit tests for each product type, compare premiums, reserves, profit, profit after solvency margin, etc., for all products before and after the merge
Perform regression tests
Document the steps taken and develop a change log along with objectives

Additional steps might be needed depending on complexity, time constraints, resistance against change or availability of resources.

Model Migration

In an insurance company with a growing book of business and an expanding product offering, it becomes imperative to have a solution that caters for the needs of the business and to achieve consistency of cashflows enabling better decision making. Most insurers consider criteria relevant for finding a solution that best meets their needs in terms of scalability, flexibility, efficiency, cost, etc.

Insurers might choose to migrate, for example:

From their existing solution to Prophet (or vice versa)
To Prophet Managed Cloud Services (PMCS) that provides computing, network, and storage resources, if already using Prophet
From asset and global libraries to the ALS library within Prophet

Migrating models may result in operational efficiency, help align models across different business units and reduce the number of manual calculations without the loss of prevalent functionalities.

Model Review and Validation

With multiple models used by firms, it has become important for insurers to have their models validated to ensure the accuracy and appropriateness of their design, data, assumptions, methodology, results, and governance. The need is exacerbated with the increased regulatory focus on processes, data quality, models, and governance.

Primary reasons for validating Prophet models include:

To validate the output from various models, including cashflow testing and to conduct model reviews on embedded value
To reduce the number of unexplained components of the model output in case of analysis of change
To compare the model output with benchmark output
To receive recommendations on end-to-end processes
To reduce duplication of efforts
To optimise and improve the performance and efficiency of the models

The validation process involves testing the cashflows by constructing, for example, an Excel spreadsheet for all the products modelled based on documented assumptions. Then, the output of the Excel model is compared to the Prophet model output for a sample of policies and any potential discrepancies in modelling are investigated and documented. Other kinds of validations include back-testing or dynamic validation.

To ensure processes are properly executed and procedures are followed, the appropriate application of model signoffs may also be tested.

Model validation forms part of the Model Governance framework which is discussed below.

Model Governance

A robust framework for model governance is essential to enforce enterprise standards. This includes:

Documenting coding best practices and standards (e.g., inclusion of code descriptions and objectives, use of input parameters)
Documentation of model changes including objectives and descriptions of changes along with impact analysis
Maintenance of testing and production environments
Incorporating and maintaining a change log including any updates for each version release
Segregating the model owner and developer while stating clear duties and responsibilities
Model version control for each release
Rules around access control

Adherence to model governance policies and standards is vital for insurance companies to ensure a smooth and reliable operation of the processes.

Conclusion

FIS ® Insurance risk suite (formerly, Prophet) offers a comprehensive solution for insurance firms enabling them to carry out actuarial modelling tasks - whether simple or complex - seamlessly. Prophet model developers use applications like DCS for data formatting and features like the diagram view and schedule to understand inter-connection between products and variables in a Prophet workspace.

Advanced features like the ALS library can help measure the impacts of various investment strategies. The use of extended formula and stochastic modelling enables firms to perform complex calculations, simulations, or scenario analyses and to better understand their results to make informed decisions.

Also, with the growing regulatory focus on model accuracy and governance, insurers may need to consider consolidating their models and enable a seamless production environment, especially when facing resource constraints. Firms may consider migrating their actuarial models involving Prophet for a range of benefits - their validation becomes an essential aspect for sound model governance and audit trail.

How can Finalyse help?

Finalyse has a dedicated team of actuaries and risk modellers who can provide advice and support for all of your modelling needs involving Prophet and beyond, including:

Development: we review, enhance, develop, or test models for both reserving and pricing purposes using best practices
Controls and Processes: we review or help improve the current control environment, governance, change control and documentation
Identify and resolve pain points: we help identify pain points or modelling challenges and resolve these to improve efficiency

References

[1] FIS is short term for Fidelity National Information Services, Inc. based out of Florida, USA

[2] FIS ® Insurance risk suite was formerly known as Prophet which is owned by FIS. Many actuaries use the terms ‘Prophet model’, ‘Prophet Professional’ or ‘Prophet workspace’ inter-changeably, we also refer to Prophet through-out this post.

[3] An accumulation in Prophet workspace allows the aggregation of results from one or more products or accumulations. It can be used to generate the input necessary for the calculation at a higher level of aggregation.

Share this article:

Table of contents

Related Articles