How Data Science Experience improves accuracy for the insurance industry
Data science is a priority for most businesses today, and data science teams are under more pressure than ever to deliver return on investment (ROI). That’s especially true, given the expense of building and maintaining data science teams. If models don’t translate to measurable business impact, efforts can be undermined.
What does it look like to deliver business value to an insurance company, where there’s a delicate balance between reducing expenses and giving customers a positive claims experience?
In this Q&A, IBM Financial Services Solution Architect Irina Saburova discusses an insurance use case with IBM Data Science Marketing Lead Rosie Pongracz. In this scenario common to the insurance industry, an organization must adjust its operations based on upcoming weather event and multiple weather indicators can improve forecast accuracy.
Using IBM Data Science Experience, a data science team can help multiple departments in an insurance provider prepare for an upcoming storm, reducing the financial and operational impact.
Finance, operations and fraud departments can benefit from the use of a diverse set of tools and techniques. In this demo video, go deeper to learn how data science helps scale predictive models and predict adjuster deployment, as well as fraudulent claims.
Rosie Pongracz: Irina, what issue is an insurance company generally dealing with in terms of business value?
Irina Saburova: A claim is the defining moment in the relationship between an insurer and its customer. It’s the chance to show that the years spent paying premiums were worth the expense. If a claim is handled well, retention rates rise. If handled poorly, the insurer may not only lose the customer, but also damage its wider reputation.
You can watch the full video here.
R.P: What are finance and planning departments facing?
I.S.: One of the most common techniques that the data science team uses in planning is called “time series analysis.” This is used to predict the number of incoming claims. Stakeholders, finance and planning can then use that information to account for financial reserves they’ll need based on our predictions. Knowing how many claims and the value of those claims is critical to determining the financial reserve requirements. Improving prediction accuracy is critical because it has a cascading effect on many other organizations within a company. In fact, our approach for improving forecast accuracy can be leveraged by almost any business regardless of industry. Later on, you’ll see how we use multiple weather indicators to improve forecast accuracy.
R.P.:Since you mentioned it, can we talk about other stakeholders within an organization that might be affected by changes?
I.S.: Certainly. The most logical area of impact is how a company deals with limited resources to address the changes. In our insurance case, it’s the claim operations group that oversees assigning claim adjusters, given the anticipated spike in the number of claims. How could this group maximize number of served clients while minimizing travel time? What is the most optimal travel route? How can they prioritize urgent cases? These are some of the questions that claim operations are dealing with.
R.P.: As far as I know, the longer claims go unattended, the higher the cost to the insurance company, so there’s obvious motivation for the claims team to start working out the case as soon as possible.
I.S.: Our solution shows how a company can optimize valuable resources; in this case, adjusters in the field. It offers recommendations in the form of schedules and maps that can be modified depending on different scenarios. This allows to serve customers faster, increasing customer satisfaction. In fact, there are many industries that use constraint optimization to address their planning and scheduling problems. For example, manufacturing, transportation, oil and gas for supply management. Or marketing campaign, portfolio and price optimizations for others.
R.P.: Are there any other groups within organization that can benefit from their data science team?
I.S.: Yes. Once claim adjusters collect and verify information about damage and its impact, the Fraud Handling Unit has to analyze the claim information. Detecting outliers and visualizing the exceptional cases is what anomaly detection algorithms are doing. These decisions ultimately inform a business model designed to minimize the costs associated with paying out fraudulent claims, as well as minimize the time spent on claims that don’t need much supervision.
R.P.: Can you bring it back to the overall business benefit?
I.S.: The main idea behind our insurance use case is about quantifying benefits and ROI for data science efforts and helping a client advance from a conceptual framework to fact-based decision making firmly grounded in data science.
R.P.: That makes a lot of sense. Our clients can infuse intelligence into their day-to-day operations and help them deliver it faster. We can serve all these groups in a unified environment such as Data Science Experience, an analytics platform that incorporate predictive and decision optimization capabilities catering to data scientists with various backgrounds and skills.
IBM Data Science Experience helps teams build and train models where they want and deploy them where they need, on premises or on the cloud. For people with different skills to work together on a project, you need a platform with tools for both coders and non-coders. The IBM data science platform ensures that models will be deployed and managed at the speed business moves today to deliver on the promise of data science and machine learning.
Ready to learn more? Get a free copy of Gartner’s latest data science and machine learning research.