Getting the Latest View of Risk Into the Business Faster
Evan CropperApril 05, 2022
Risk modeling teams experience constant pressure to ensure their business uses the latest view of risk. In an ideal world, risk models across all perils and regions would be validated and incorporated into the business as soon as they become available. But in the real world, risk model validation is a long and intensive, but necessary process.
Many model vendors including RMS® release an annual major model build as a single “all-or-nothing” on-premises software package.* A typical release will include many new and updated models, upgrades to geocoding, as well as updates to the modeling software itself. To complete the validation, users need to validate the whole version or run multiple production environments.
Because of the way catastrophe risk model software is packaged, (re)insurers and brokers often face a difficult decision on how to incorporate updates into their workflows. Should they validate all models in the release together – sacrificing speed to market – or should they prioritize to get the most important models to market faster without conducting the same due diligence on the other models? Both options bring risks to a (re)insurer, such as exposing the business to greater portfolio losses.
When Software Packaging Impacts Model Validation
The best way to understand the model validation dilemma is to explore a real-world example. In 2017, RMS released Version 17 of the RiskLink® software that included multiple significant updates to the North America Earthquake, North Atlantic Hurricane, and Southeast Asia Earthquake Models.
For many of our clients, all these models were of high materiality and therefore required scrutiny and a lengthy validation process in order to adopt. And like any other release, our customers dedicated significant IT and catastrophe risk modeling resources to the extensive model data runs required for validation.
Because all of these clients ran Version 17 using on-premises systems, they may have also experienced common challenges to validate and adopt a package of new models, such as IT bottlenecks and data quality issues. (I have explored both these issues in previous blogs; see links in the “You May Also Like” sidebar.)
For our broker clients, this could have been particularly challenging to the business because having the most up-to-date view of risk is critical to best serve their clients.
Rethinking Model Validation Unconstrained by Software Packages
The RMS Risk Modeler™ available on the RMS Intelligent Risk Platform™, can help (re)insurers rethink their approach to validation and adoption for new models – even entire version packages. Just a few of the ways that (re)insurers benefit from the cloud-native SaaS environment include:
When a new version is released, all models within the package are uploaded and ready to use in the Risk Modeler application. IT does not need to set up a development environment for model validation.
The Risk Modeler application decouples the models from the software packaging, allowing a cat modeling team to pick and choose which models to operationalize based on their materiality to the business.
All versions of the model run on the same exposure data. The cat modeling team can understand changes in the model faster because it is confident that the models share the same underlying data.
Finally, in rare cases that a (re)insurer prefers an older version of a model, it can choose not to migrate to the latest release without impacting the validation and adoption of other models packaged in the release.
Looking at Version 17 of the RiskLink application, using the Risk Modeler application a (re)insurer could validate the U.S. Earthquake Model first and adopt the model for use in production before the renewal season (Figure 1). The business could then move to the U.S. Windstorm Model or the Southeast Asia Earthquake Model later, depending on business priorities or materiality, without delaying adoption of the U.S. Earthquake Model.
In 2021, RMS made Version 21 generally available to all our clients running either the RiskLink or Risk Modeler applications; this was the first time the release was available simultaneously on both platforms.
With the general availability of Version 22 just around the corner in late June – with significant updates to the China Earthquake, U.S. Winterstorm, and Central and South America Earthquake Models – many of our Risk Modeler clients will benefit from a faster and more efficient model validation process.
In some cases, we have seen customers reduce the validation process from months to weeks. More importantly, the cat modeling teams will be empowered to prioritize model validation based on a model’s importance to the business, reducing any friction between modelers and the wider organization.
*RMS High Definition Models™, including the U.S. Inland Flood HD Model and U.S. Wildfire HD Model, are not restricted by annual software releases. Model releases, updates, and upgrades are driven by customer priorities and market dynamics.
Evan leads climate change and modeling product marketing for Moody's, where he helps customers develop more data-driven strategies using physical risk analytics. He has extensive experience scaling technology in the digital enterprise with a passion for using data to deliver better business outcomes.
Previously, Evan worked in various product management and marketing roles with Hitachi Vantara, Current - a subsidiary of GE Digital, and Cisco.
He holds a bachelor's degree in Political Science from Emory University and an MBA from Vanderbilt University’s Owen Graduate School of Business.