Getting ahead of the curve on data quality

The goal

Incomplete or inaccurate data assessment can bias model results. In Moody’s latest Illustrate case study, we look at how Moody’s worked with Securis to enhance the quality of its cedant-provided data.

Securis Investment Partners, a leading insurance-linked securities (ILS) manager, is committed to ensuring best-in-class analytics throughout the business.

As part of this, the private non-life origination team wanted to understand how the latest exposure data augmentation tools and techniques could be used to enhance the quality of cedants’ underlying exposure data and what impact this could have on their risk selection and pricing strategies.

“Catastrophe model output is still the common language of risk for risk transfer,” says Paul Wilson, Head of Non-Life Analytics at Securis. “This was an opportunity to explore the value of assessing and potentially investing in augmenting the data we receive.”
 

The objective

In 2019, Securis commissioned Moody’s consulting services to analyze and enhance some of its cedant data and come up with practical recommendations for improving its data assessment processes.
 

The process

Data quality assessment

To understand Securis’s cedant data, Moody’s ran a series of data quality analytics across all portfolios to assess the data for both completeness and accuracy. Moody’s data quality analytics are informed by the catastrophe models and score the data based on the impact that unknown or inaccurate data could have on modeling results, providing greater insight than more basic metrics.

“If you are looking at the exposure data for running a hurricane model, the locations on the coast in Florida will have a far greater impact on modeled losses than inland locations, so you need to have data quality scores that account for this,” says Edwina Lister, Lead Consultant for the project. “Similarly, you need to take into account which attributes drive the model results”

Accuracy is harder to assess than completeness but inaccurate data can introduce far greater bias in results. Moody’s assesses accuracy in a few different ways; this includes running a set of >120 validation theuristics and comparing individual locations to the Moody’s exposure source database.

“It is easy for a cedant to improve completeness at the expense of accuracy, and this can introduce material bias into the portfolio” says Will Mayes, Head of Business Consulting at Moody’s.
 

Data enhancement and deep dive

The project team worked with Securis to identify cedants using Moody’s data enhancement engine and through manual investigation. The aim was to achieve the best possible representation of the selected exposure sets and understand the change in loss associated with each update. “This helped us understand the materiality of the data quality challenges and informed our recommendations to enhance Securis’s process,” says Lister.
 

The solution

Recommendation formation

Moody’s consulting services and Securis worked to develop pragmatic recommendations that could be implemented within Securis’s underwriting workflow. These recommendations leveraged the Moody’s Risk Maturity Benchmarking framework and defined both short-term and longer-term objectives.

“We provided detailed recommendations to help Securis improve its data assessment process and developed a data quality loading framework based on the cedant analysis. This allows Securis to have greater insight into the model results and make better-informed investment decisions,” explains Mayes.

The outcome

Client perspective

It was important to have senior buy-in for the project to be a success, and this came from Securis’s Chief Underwriting Officer Paul Larrett and throughout its origination and analytics team.

“We do expect to get some real value from this, and it will change how we approach some specific investment opportunities,” Larrett concludes. “This project has given us some interesting insights into the potential of data augmentation and how small changes in our workflow can make a positive difference.”