Overcoming the Predictive Modeling Stumbling Blocks in Commercial Insurance

Mirroring the four-step process of a product development lifecycle provides a best practices blueprint for overcoming the many obstacles associated with implementing predictive modeling in small commercial insurance.

(Image credit: Adobe Stock.)

Predictive modeling has become an invaluable risk assessment tool in personal lines insurance. It has enabled insurers to ascertain risk and purchasing behavior data to determine pricing far more precisely and profitably.

Yet despite the many similarities between personal and small commercial insurance, predictive modeling has been slow to catch on in the latter segment. Most small commercial underwriters understand the huge potential, but often balk when it comes to implementation. They worry that the company’s leadership may not be supportive. Or they’re unsure whether enough data exists to build a model, or uncertain about which data sources to tap. Or they worry the company lacks the necessary in-house expertise and resources to justify the investment, among other stumbling blocks.

To overcome the paralysis, a few best practices can make implementing predictive models achievable for any carrier, regardless of their expertise level. The following is a simple yet proven best practices framework for integrating predictive modeling to reduce your company’s risk exposure and grow your business. What’s interesting is that this four-step framework closely mirrors the four stages of the product development lifecycle: ideation, design and development, implementation, and monitoring.

Step 1: Ideation
The success of any predictive modeling initiative requires everyone involved to have skin the game. It requires strong executive sponsorship to ensure all the right resources will be applied, and a it requires a committed cross-functional team to bring the idea to reality.

In the ideation phase, the team begins by showcasing the benefits of predictive modeling and establishing buy-in across the organization. They must identify and prioritize the key problems to solve through predictive modeling, determine the cost and ROI of the models to be implemented and figure out how to integrate the predictive model into the underwriting workflow, including measurable success benchmarks.

Step 2: Design and development

Predictive models are useful for risk selection, pricing, claims fraud detection, claims subrogation potential and other purposes. But within small commercial predictive modeling skews toward risk assessment and pricing by creating insurance scores that rank risks based on loss propensity. Designing this type of model is an iterative process, beginning with data exploration, the creating the model, and finally, ensuring it complies with any applicable regulatory requirements.

The data exploration requires a team of business analysts, statistical modelers, IT resources and regulatory experts. Third-party data sources can be tapped for building a predictive model for risk assessment – including commercial credit, consumer credit, and public records.

Risk assessment requires a large amount of premium and loss data for training the model to predict the target, as well as to test that it works. Data is partitioned as either training data or testing data for the project. Of course, all data sources and attributes used to predict the target must comply with any applicable regulatory requirements. For insurance companies that don’t have sufficient data to build their own model or the necessary expertise, a model developed by a vendor or consulting group is a viable option.

Step 3: Implementation

Once the model has been designed and proven, it’s ready to be implemented within the workflow. Because implementation impacts so many parts of the operation, the team needs to identify and document the impact to existing business rules and procedures, such as rating and underwriting.

They also must determine the IT requirements for building the model, application workflow changes, and score storage and tracking.

Other requirements for implementation include making sure that any applicable customer dispute process is supported. Training all stakeholders and impacted parties comes next. And then the team creates a rollout plan.

Step 4: Monitoring

With all the hard work completed, the last step is ensuring your model works as designed. Monitoring lets you know if the model is meeting performance expectations.

There are two key parameters to monitor: usage tracking and model efficacy.

When using a predictive model for insurance scoring, teams will want to know when scores are being used and when they are being overridden. Both tallies should be tracked, and the who and why behind the overrides reported out. Score overrides can provide valuable insights into any score limitations, how the scores are adopted, and opportunities for improvement.

Monitoring for model efficacy reveals if the model is meeting performance expectations. If it’s not, a deeper dive into the underlying causes is needed. Sometimes all it takes is a minor recalibration. More serious cases may require rebuilding the model, or creating an entirely new model based on different assumptions and data sources.

Final thoughts

Embracing predictive modeling can be intimidating for small commercial insurers because there are so many moving parts, diverse constituencies, and often a lack of confidence in the ability to do the job right. Mirroring the four-step process of a product development lifecycle provides a best practices blueprint for overcoming the many obstacles. Once you’ve integrated predictive modeling into the workflow, insurers will be more successful in protecting and growing their book of business.

Applied Insurance Analytics: Q&A with Author Pat Saporito

Mathew Stordy // Mathew Stordy is Director of Commercial Insurance for LexisNexis Risk Solutions. Stordy is responsible for requirements assessments and the design of data solutions and services that streamline commercial insurance processes and provide insights about entities through the use of data, analytics, and software. He has more than 20 years of experience focused on insurance software and specializing in property and casualty insurance systems. Stordy has worked in all phases of the systems-development lifecycle. Prior to joining LexisNexis, he worked extensively with policy administration systems, quoting applications, and business intelligence solutions. He holds a bachelor's degree in mathematics with a minor in computer science and a master’s degree in philosophy from the University of Connecticut.

Leave a Comment