How the insurance industry can make sense of big data

By Brian T. Horowitz

When car insurance companies accumulate more than 100,000 records of insurance claims — and many of them do — they need a platform to make sense of the data.

Organizations such as Netherlands-based Delta Lloyd encounter lots of unstructured data, including information on vehicles, drivers, policyholders and repair shops. Unstructured data often consists of unorganized text that software programs are unable to fit into set fields in a database.

The deluge of unstructured data in insurance comprises claim details and policy descriptions — “anything longer than three words,” noted Jan-Kees Buenen, CEO of SynerScope, a Netherlands-based company that offers a comprehensive data analysis platform.

SynerScope’s software provides a multidimensional view into sales and claims data. It enables domain experts, those who analyze data from sales and claims handling, to make decisions on whether to accept a new policyholder or marketers to decide whether to launch a new campaign.

With its software, SynerScope aims to shorten the time required for domain experts to analyze reports.

Finding patterns in insurance data

“For Delta Lloyd, the biggest value is to find structures and patterns in data that were unknown before, or impossible to be found using traditional analytic platforms,” Peter Schaap, manager of information management for Delta Lloyd, told Power More in an email. “With SynerScope we can do analysis in a powerful intuitive way, without the limitations of old.”

In the past, multiple workers were required to carry out a full analytic cycle from data to insight, Schaap recalled.

The SynerScope server-side platform brings more efficient analytic processes, he said.

“In addition, the platform offers a unique opportunity to allow people with less formal data science or analytics education to perform high-quality data analytics,” Schaap said. “Their different perspective on topics for analysis sometimes produces surprising results.”

Advanced scientific visualizations

Dell OEM Solutions built a proof of concept to determine the hardware that could work with SynerScope’s visual analytics software suite. The platform Dell provided incorporates Intel Xeon-powered PowerEdge servers as well as Nvidia Grid K1 and K2 cards, which enable rich graphics in virtualized environments.The Nvidia cards allow the software to offer advanced scientific visualization as 3-D objects, which get transformed into a topographical representation. The GPU-accelerated servers enable SynerScope to bring information processing, or compute, close to where the data is being collected.

Server-side virtualization provides flexibility in the type of data sources and data types companies can access. When combined with high-speed analytics, server-side virtualization brings short cycles from loading data to first insights, Buenen said.

“A server-side implementation allows us to make the capabilities available throughout the organization in a fully efficient manner,” Schaap added.

The SynerScope platform also supports BYOD and allowing employees to work from home, Schaap noted.

A single-tool environment

Companies such as insurance companies can make sense of big data by using a single tool environment. Within one interactive interface, SynerScope combines four dimensions, or sources of data: text, sensors, video and numbers.

“So now you can actually choose the dimension from which you want to look at the data,” Buenen said. “You have the option as a user to look from almost any dimension that is possible on the data on the fly without having to go back to IT or a data scientist.”

The company’s software uses telematics along with GPS transmitters to help with data analysis for car insurance and leasing. Insurance agents can feed images and video into the database from Google Street View to perform deep learning and get insight on objects in a picture. Agents can search for info on all the roadside accidents in an area to help evaluate insurance claims.

If information comes up on search criteria — say people in a certain age bracket that have been in accidents — the visual analytics software will show a group of orange dots. Larger dots for a specific topic mean a larger sentiment, or problem area.

The image-recognition capabilities are a combination of deep learning and deep belief networks, which are graphical models that show connections between layers rather than units within each layer.

SynerScope’s platform also combines network analytics, text analytics and geography analytics. It differs from traditional data mining because it uses human perception to analyze visual cues the software presents.

The application magnifies the amount of data the brain can digest, Buenen said.

A goal is to find a connection between the data and its context, such as a picture of a damaged car and its surroundings, or a church and meadows around it.

Another important aspect of data discovery is the software’s ability to skip the data modeling step, which is very time-consuming. With the SynerScope platform, users can avoid going through all the data cleaning and modeling that often comes before they can make inferences from the data, according to Buenen.

“That makes it ideal for data discovery,” he said.

“Between 60 and 80 percent of the time in data analytics is actually spent on curating the data, cleaning it, modeling it and indexing it, and only then can the process start,” Buenen said.

With all the data that companies need to process, the goal for organizations is to access the data, map data relationships over time and gain insight as fast as possible.

“Big data can only help us if we can derive meaning from it,” Buenen said. “Somebody has to be able to understand it.”

About the Author: Power More