Reserving is a prime example of how process industrialisation can significantly improve actuarial practices and the value of actuaries in business decision making.
If you haven’t already, find out why process industrialisation is critical to your success. Then read on for our reserving specific, Psicle use case.
Our clients are rapidly rejecting their legacy reserving processes. Built in fragmented environments, these processes lead to fragile, resource intensive and slow modelling. Key person risk soars and new joiners struggle to assimilate the maze of data flows.
Given the complexity and randomness of the claims process, reserve estimates are inherently uncertain – fragile reserving processes only amplify uncertainty; resulting in mistrust and ill-informed decision making.
It is not enough for Boards to be simply presented with the actuarial “best estimate” view of reserves. To make informed decisions, actuaries and Boards need to understand the assumptions and methods relied upon and the sensitivity and volatility in the reserves.
It is crucial that the industry assesses “reserving” across IFRS4/GAAP, Solvency II and now IFRS17. Psicle has the ability to reserve, cashflow, discount and simulate, which allows you to model over all three of the regimes within the same environment. This brings a step-change in the way the industry is overcoming the challenges that lie with embedding these business-critical models.
Reserving in Psicle is a paradigm-shift from traditional reserving approaches. With Psicle, blind reserve results are available within minutes, with the actuary, by exception, able to tweak the assumptions and select different methodologies within interactive mode. Actuaries’ time is dedicated to applying expert judgement, comparing a library of methods and running a suite of sensitivity tests. Our Reserving functionality has a number of pre-built reserving methods, including the standard approaches used in the London Market, as well as a growing suite of international techniques.
- Single environment
Reserving actuaries are used to disjointed processes; relying on multiple, yet incompatible, tools and creating significant side projects in Excel to overcome deficiencies.
Triangle reserving often has data extraction, cleaning, and manipulation in manual processes using different tools, then reserved elsewhere (often in Excel or often in third party software with heavy Excel reliance to perform key calculations) and pushed to data stores. The downstream links to SII TPs and SF, and now IFRS17, are additional manual steps in the process.
With Psicle, the full reserving process (and beyond – see our SII TPs, SF and IFRS17 use cases) sits in one robust environment, even on the same canvas. This is a step change in the way we, as an industry, currently tackle our problems.
Our reserving gadget supports the standard methods – such as the Initial Expected Loss Ratio (IELR), Chain Ladder (CL) and Bornhuetter-Ferguson (BF) – the Munich Chain Ladder – which is popular with our Scandinavian clients – and a range of Australian methods. These include the Projected Claims Estimate (PCE), Payment Per Claim Finalised (PPCF), PPCF in Operational Time and Payment Per Claim Incurred (PPCI). Psicle also has reserve uncertainty capability, including the Mack method and Bootstrapping.
Methods have been incorporated over time in response to increased sophistication in the market. Process industrialisation takes away the laborious data manipulation tasks from analysts and actuaries, creating space for technical curiosity and exploration of emerging methods.
- Dynamic parameters
Assumptions are dynamic, feeding off the prior period’s results or from side analysis built within the project. For example:
- automatically use last period’s ULRs as this period’s BF IELRs; or
- let Psicle determine the appropriate large / non-large claim split.
Sensitivity tests are run automatically using scenarios pre-defined as alterations to the base case. For example, take all base selections (assumptions and method blends) but:
- take the CL pattern to be the average of the past 10 periods; or
- apply explicit claims inflation.
Within minutes of finalising the base results, the actuary can understand the materiality of assumptions and the potential volatility in results, enabling evidence-based challenge of the selections. Psicle’s unique view on how a reserving process should operate enables our clients access to much broader range of information significantly faster than traditional approaches allow.
You don’t need to be a coder to build or enhance projects in Psicle. Manipulations rely on easy to use wizards that are powered by the SQL language and the reserving gadget itself is instructed by one fully configurable parameter (or “instruction”) file.
It is simple to increase granularity if you have the data available. For example, easily switch between class or sub-class level and between monthly or quarterly development periods.
We work collaboratively with our clients to build the initial reserving project, replicating standard approaches, but the majority of our clients have built bespoke side analysis and reporting processes over time.
- User interface
Psicle’s reserving gadget is highly visual and fully interactive, making it easy to: exclude triangle entries, override patterns, alter initial parameters and blend methods. Actual vs Expected graphics provide a sense check on the selections.
Reports can be produced outside of the reserving gadget using Psicle’s Reporting Object or results can automatically populate a specified Excel template.
We have recently enhanced our functionality to automatically produce an “AoS” (Analysis of Surplus) which is an extremely onerous regulatory requirement in Australia. The AoS will automatically produce a waterfall chart of the drivers of change in the reserves from period to period, where the user can configure exactly which drivers to include.
A key feature of Psicle is its connectors to external data sources, to automatically pull and push data as part of the model run. We have built connectors to SQL databases, flat files, Excel spreadsheets and generic web API’s; with a variety of other connectors currently in development. This drastically reduces the time taken to run the reserving process and the potential for data errors. Through the use of PsiCode (our proprietary pseudo language) Psicle can invoke calculations in other languages such as Wolfram Mathematica and Azure Machine Learning.
Click here to read reserving case studies, or contact us for a demo!