Challenge

In the insurance industry, there are many factors which impact the level of risk posed to each insurer. Quality Planning Corporation ("QPC") wanted to be able to identify and correlate a broad range of historical information in order to improve the way insurance companies currently rate auto insurance policies.

Risk factors can be location-based or not. Correlating potential risk factors with actual risk required QPC to analyze a number of historic datasets. QPC suspected that the location of a policy holder in relation to several risk factors might lower or raise the risk of an automobile accident. Understanding the components affecting risk, particularly with respect to a policy holder’s location, means recognizing and analyzing potential factors based upon their locations and surrounding conditions.

Once QPC identified a set of potential risk factors, large volumes of data had to be compiled and compared against the location of historical policy records. The predictive nature of the potential risk factors could then be assessed statistically. Although insurance ratings have traditionally relied on zip codes as a means of identifying the location of policy holders, QPC required much more accurate location-based data to analyze potential risk factors.

QPC was faced with the challenge of efficiently geoprocessing massive unstructured datasets (totaling more than 300,000,000 locations) and developing a highly flexible, customized and configurable nationwide set of potential risk factors, all within a limited budget and timeframe.

Solution

PC engaged Farallon Geographics to identify and correlate a broad range of location-based information that could be statistically analyzed to determine whether specific spatial relationships between risk factors and the location of policy holders correlate with levels of risk.

Oracle Database 10g, with the Oracle Spatial and Partitioning options, was selected as the optimal technology platform because it offered:

• Superior scalability
• Ability to efficiently process hundreds of millions of records
• Native spatial data type supporting standard SQL access and access through published PL/SQL and Java APIs
• Open interoperability across all major GIS platforms
• Complete server side geoprocessing

Farallon Geographics maximized geoprocessing performance by using the diverse standard and spatial optimization features of Oracle 10g. Since the project data were static, a data warehouse approach was applied. Data were organized into a series of structures tuned to support all required geoprocessing. For example, Oracle 10g’s extensive spatial and tabular partitioning optimization was leveraged and tuned to the specific geoprocessing requirements. With this solution, QPC will be able to geoprocess tens of millions of individual policy locations very rapidly.

The Oracle Database 10g platform with Oracle Spatial allowed for massive data processing capabilities and scalability within an open and configurable architecture. With training from Farallon Geographics, QPC will be able to maintain the capability in-house without significant additional costs.

Farallon has enabled QPC to define a completely customized set of potential location-based risk factors, and to compile those factors into nationwide data sets to be analyzed against historical claims and fraud to isolate the most relevant factors. This provided the flexibility to broaden the type of risk factors considered, resulting in large improvements in the accessing the risk of various types of insurance policies.

Categories: Spatial Data Processing, Risk Management, Planning & Training, GIS Needs Asssessment, Training & Technology Transfer

Selected screenshots from the project

Click on any thumbnail to view a high-res image

Geoprocessing Overview Schematic

Proximity zones