Insurance

Customer Opportunities & Challenges

Across the many sub-segments of insurance spanning commercial and personal, including auto, home, renters, life, disability, real estate, etc., there are opportunities to do better risk modeling by combining diverse sources of data to improve the fidelity and specificity of models, and to better anticipate emerging risks. These high variety sources might include video imagery, such as overhead views of a building’s condition or parking lot occupancy, sensor data such as might be collected from tracking systems within a vehicle, and even external sources such as social media data, news feeds, or weather data. If this data could be effectively used together, it would allow companies to better serve customers through more competitive premium pricing and timelier servicing of claims, while better managing their own exposure and reserves.  Trends toward online interfaces that offer customers quick response times push attendant requirements for real-time analytics and automated decision making.

Among the challenges insurance customers face in realizing these opportunities are: 

• Managing cost, complexity, and analysis latency with information systems that combine diverse sources of data
• Establishing what sources and in what combination provide the most significant benefits in terms of improved risk modeling of specific event categories
• Dealing with issues of “mixed staleness” in data sources because modeling efforts cannot keep pace with dynamically changing source data
• Analyzing risk in real time, and/or monitoring it continuously, based on the most current information available

Our Value Proposition for Insurance

Collaborative Analytics (CA) provides cross-silo risk assessment which is agile by design, since the data are never integrated, and the model learning and prediction components are fully distributed processes. This leads to the following benefits for insurance customers:

• Risk analysis capability that is flexible to combine distributed and silo’ed information in a way that is cost effective compared to schemes that integrate data, scalable to new sources incrementally, and can take advantage of sources even when the raw data cannot be accessed; customers can make use of the most sources of data and the latest data available
• An agile modeling capability that supports rapid model update as the dynamics of sources change, with an ability to update the global data model incrementally as the system operates
• Ability for companies to use best-of-breed learning and modeling techniques tailored to the specifics of the data in a silo, without having to deal with understanding and reconciling to other silos
• Mitigation of governance, access, and redundancy concerns around large centralized pools of data
• Real-time decision making and continuous monitoring of risk at a rate that is not possible with schemes that have to integrate data and re-learn centralized models

Because the insurance industry has a long history of modeling risk, most companies already have some kind of operational capability in place. A common scenario is that existing models are “expert system” or rule-based systems that are supported with manual intervention and oversight to “fuse” in other information and considerations, while internal development effort is being expended to build more automated capability based on machine learning and the latest methods in artificial intelligence.  A unique value proposition of CA is that it can augment existing models with additional sources of information, mined with the latest techniques, without opening up the core model in use or even understanding its inner (proprietary) workings. This provides an extremely low touch way to try combining additional information to assess the benefit without disrupting the current approach in use.                       

CA provides an agile cross-silo analysis capability to assess and monitor evolving patterns of risk for insurers across dynamic distributed sources of data.