![](https://edge.sitecorecloud.io/guidewiresodb06-guidewire0f2e-productioncd91-5186/media/project/guidewire/guidewire/blog/thumbnails/blog-image-20230228-two-at-computer-monitor-1200w.jpeg?h=630&iar=0&w=1200)
Many insurers have experience implementing advanced predictive models in their pricing and automotive lines of business. Despite the proven impact of these models, some insurers are missing the boat when it comes to exploring new opportunities.
When predictive modeling was new to the industry, significant resources were allocated towards hiring analytics experts and data scientists to create these models. In many cases, senior management recognized that this was not their area of expertise, and internal R&D teams were formed to develop and deploy these predictive models via predictive analytics.
However, over time, R&D teams at many organizations became disconnected from senior management. Insurance business leaders often deferred to the experts in R&D when it came to predictive modeling, which proved to be a mistake. The priorities of R&D teams today are all too often misaligned with the vision and priorities of leadership.
Data science professionals excel in creating predictive models, but they may not be as effective in applying and monitoring these models on a day-to-day basis. Additionally, they may not have the same level of understanding of overall business challenges and opportunities as senior management. Identifying business challenges and opportunities, and developing and applying predictive models, each requires a different skill set.
Efficient organizations apply the appropriate skills to each part of the various problems. Business leaders should define the business need. Data scientists and actuaries should scope out the problem, defining what data and predictive models can provide the information which will allow the business to address the problem. IT professionals should collaborate with the modelers to implement models, and business analysts should monitor and report on the end results.
When a carrier decides to develop a predictive model to tackle a business issue or exploit an opportunity, it's crucial to approach the project with a holistic business perspective and recognize it as a continuous effort. Carriers must remember that predictive models are a means to address business problems, not an end solution in themselves.
To advance your predictive modeling program, we recommend you take these steps when addressing a new challenge:
Tackle a significant business problem - To develop an effective predictive model, consider the valuable data scientists and resources required. Investing in these experts and resources wisely, ideally by applying them to the scoping of the problem, gathering needed data, and creating the actual model – and that these are done in furtherance of addressing a significant business problem.
Have a clear plan of action - Understand the plan of action before starting to develop the predictive model. While there are standard approaches and tools available, the actual application and management of such models should not be an afterthought. Understand how the model will be implemented and how data will be fed into it in the field. Integration of predictive model results into core systems should also be considered to make the insights actionable.
Continuously monitor performance – Once implemented, monitor the model’s performance continually. It is unreasonable to assume that a predictive model will work exactly as expected once you flip the switch. Unintended developments and consequences are normal in all business undertakings, including predictive model applications. Continuous improvement is necessary to maintain an effective model, particularly as your business and market conditions evolve.
Adjust the predictive model - Be prepared to adjust the predictive model. Monitor results and obtain feedback from the field to determine when and where adjustments may be needed.
Know when to involve data scientists - Eventually, it may be necessary to bring data scientists back into the picture to improve the approach taken. Knowing when models are working optimally and when they're not, and when a business problem is significant enough, is essential. Bringing data scientists back in to exercise their specific skills can substantially improve outcomes.
To solve a problem holistically, carriers should focus on the business problem rather than just the predictive model. By honing the development and execution of predictive models, carriers can continuously improve insurance processes and performance, providing a significant competitive advantage.