Professional Insight

Insurers Enjoy Benefits from Data Modeling the Claims Process

Applying data analytics to the claims process is reaping a multitude of benefits and future opportunities, panelists agreed at the CAS Spring Meeting session, “Data Science and Improving Claims Customer Experience.”

Marty Ellingsworth, senior analyst at Celent and moderator of the May 12 session, began by explaining that improving claims processes through data science helps enhance the customer experience, which is one step in the overall customer relationship journey with the insurer.

Improving the claims process through analytics results in lower costs, better quality and consistency, faster processing and an enhanced customer experience, says Eric Sanders, head of claims for QBE North America. “We’ve already seen all this happening, and it’s really exciting,” he adds. Thanks to data analytics, the panel agreed, there has been more improvement to the claims process in the past two years than during the prior few decades.

One breakthrough is using predictive models to triage the claims process, ensuring that applications go to the appropriate adjuster, says Tom Warden, senior vice president and chief data and analytics officer for Employers Insurance Services. His company is deploying automation to make more efficient decisions and to minimize the time that claim reps and adjusters spend on routine decisions. “We are trying to focus adjusters on the decisions that really require their intelligence,” Warden explains. In turn, this hastens meeting the workers’ compensation goals to encourage the healing of injured workers and their return to work.

Thanks to data analytics, the panel agreed, there has been more improvement to the claims process in the past two years than during the prior few decades.


By using “some pretty sophisticated data science, [the company is] continuing to refine our approach based on the more complex modeling and the feedback we get from our adjusters,” Warden offers. This helps staff find claims with a high potential to become “jumper claims” that start off with $10,000 or $15,000 losses and can jump to $200,000 or $300,000 losses due to attorney involvement.

To accomplish this, “We’re taking the decision processes that our best adjusters use and putting that in the code and, in essence, creating complex business rules that address specific situations,” Warden explains. “The data science really comes in on the back end,” he adds, which allows learning continuously from decisions and trying to improve upon business rules written in the code. “So, a lot of machine learning will be used to modify and update the business rules in the models being used in the day-to-day process.”

USAA is aiming to allow customers to self-service less complex claims, says Luke Harris, the insurer’s assistant vice president of innovation. “We want [adjusters] focused on that 25% of work that is the most complex, whether it’s from an empathy perspective, or whether it’s from a truly complex type of claim process or claim event,” he explains. The personal auto and home insurer has set a goal to automate up to 75% of claims “without material impact to the member experience or the employee experience by 2022,” when the company is celebrating its 100-year anniversary. “We believe it is attainable,” Harris adds.

Emphasizing that relationships still matter, QBE’s Sanders offers that his company is using “data science to gain insights but also to drive relationships by way of how we automate claims, and how we use the data science in a smart way to improve the customer experience.” At the same time, data is being used to improve upon other areas, such as loss prevention. Since QBE has a diverse professional lines book of business — including directors & officers, errors & omissions, transactional liability, trade credit, surety and more — the insurer’s goal is to automate about 25% to 30% of claims within three years.

Data analytics is also effective for fraud detection, USAA’s Harris observes, but proper controls are necessary to accurately pay a claim instantly while still catching fraud. “Identification of fraud certainly is a byproduct of having the proper controls and very sophisticated models to identify claims that deserve to be paid,” he offers. “I really see a shift in approach where it’s less about catching the fraud [and] it’s more about identifying the claims where there are elements of fraud, and then very quickly turning those, with the byproduct [that] the bad actors do get caught.”

While there are injured workers who try to remain on benefits as long as possible in workers’ compensation, Warden sees “the real fraud” coming from doctors and lawyers. Thanks to predictive modeling, Employers Insurance Services is successfully finding organized fraud committed by medical provider networks and bringing it to district attorneys for successful prosecution.

The potential for data analytics in the claims process has not yet been fully realized. “The biggest untapped part of the data and analytics equation is around how claims can really support underwriting in informing intelligently how to underwrite business,” Sanders says. He also sees opportunities to address the impact of litigation, which requires large-scale data.

Panelists agreed that ensuring a significant amount of quality data is a central challenge while maximizing the potential for applying analytics to the claim process. “You need scale, or you need external data to supplement internal data,” Sanders says. “You need your claims team coding accurately. That’s not something that anyone should take for granted,” he adds.

The potential for data analytics in the claims process has not yet been fully realized.


For workers’ compensation, Warden says, part of the difficulty is that claims adjusters are under pressure to move quickly. Sometimes they have to skip steps on claims systems just to get the claims open, which does not lend itself to data quality. “All our automation efforts and our modeling efforts are really based on that source data,” Warden says. His team is partnering with claims adjusters to encourage efficacy and ensure that data quality is “as pristine as it can be.”

USAA’s Harris says it is important to determine how to prioritize, pursue or collect data in order to solve business problems. He suggests examining the cost of incremental enhancements or improvements and considering whether there may be a lower-fidelity solution to that same business problem.

More data and analytics possibilities abound as well. QBE is now using tech tools to address claim-related challenges. The company is running a pilot project for fraud detection that examines digital picture data to detect manipulated photos. “Actuaries are going from loss triangles to predictive modeling to internet of things data sources, such as using geospatial data with voice and video and collusion networks,” Sanders says.

Sanders also sees data analytics and tech tools as ways to help enhance customer experience. “In our crop area, we’re using microclimate weather data not only as a method of managing a loss better, but [also] to round out services we provide to the American farmer to help them manage their crops better,” he says. “You’ll see a lot more of that going forward where carriers will certainly be interested in rounding out the types of services that they provide.”

Annmarie Geddes Baribeau has been covering insurance and actuarial topics for nearly 30 years. Find her blog at