Putting predictive analytics to work at Infinity Insurance

I am giving a tutorial and a presentation on putting predictive analytics to work at the forthcoming Predictive Analytics World show in Washington DC (October 20th-21st with tutorials on the 19th). I always like to illustrate my points with real examples and while I was preparing the presentation I had a chance to talk to Infinity Insurance about their experience putting predictive analytics to work. Infinity had some great results when they put predictive analytics to work in their claims operations, including:

1100% increase in fast track rate
33% higher returns subrogation returns
Analytic software investment paid off in 3 months
Subrogation recovery up by $10m/year

1100% increase in fast track rate
33% higher returns subrogation returns
Analytic software investment paid off in 3 months
Subrogation recovery up by $10m/year

Infinity Insurance is a $1Bn writer of insurance for classic cars, commercial auto and personal auto. The claims department at Infinity signed up with predictive analytics software from SPSS Inc. in 2007 to target process change in a couple of areas – fraud and subrogation. The Fraud Investigation Unit had an old process for flagging potentially fraudulent claims, involving adjustors manually applying some simple rules. This needed to be replaced with a more transparent system to identify risky claims. In Subrogation, the process of finding other companies who might be liable for some of a claim, Infinity had spent millions of dollars for outside consultants to come in and check for subrogation opportunities. This had yielded $12M – $16M in actual recoveries so it was clear that a better process was possible.

READ

Predictive Analytics And Fiat Currency Make ICOs Excellent Investments

Subrogation turned out to be the quickest payoff, with analytically-derived rules being deployed in about 9 months. The new subrogation decision had 33% higher returns – it found subrogation opportunities in many more claims than before – and paid off the analytic software investment in just 3 months. Previously Infinity had about 15% of its claims come in to the subrogation department for review and now over 22% come in. As a result recovery from has gone from $1.2m/mo to $2m/mo!

While Infinity was adopting analytic decision making the industry entered a deep recession. Without the analytically enhanced decision-making they were adopting this would have had far more effect on their business than it did. They had to lay off 25% (some 300 staff) and trim expenses. With 12,000-13,000 new claims a month they had to get cases out to field staff quicker, improve customer service and reduce people all at once. The automated identification of fast track claims helped them decouple the number of adjusters from the amount of business being written.

To achieve this they started “Right tracking” claims using Risk Control Builder to assign claims to fast track adjusters. This was a new approach, created to take advantage of the new analytic decision making using a combination of business rules, predictive models, and information gathered from customer interactions. These fast track adjusters were people in the business units who went from simply reporting losses to handling some claims. This created a group who could process the claims “once and done” – no hand off to field adjusters. This allows them to open, appraise and pay some claims within 10 days or so. The analytically enhanced decision increased the old fast track rate of 2% to 22% in just a year of operation. The 100 fast track adjusters now handle more than 2,500 claims a month without referral to a field adjuster. This represents a huge cost saving for the field, helps decouple business growth from the number of field adjusters and has reduced their loss adjustment expenses from 14% to around 11%.

Infinity uses PASW Modeler data mining workbench to create predictive models and Risk Control Builder to deploy them. Starting originally with external consultants, Infinity has been transitioning to having more of its own staff use the tools. The resulting decisions are a mix of rules specification and analytics work and Infinity has been identifying people who have real claims experience and getting them involved in managing the rules themselves as analysts. This enables new rules, new models to be deployed without going through IT and this is important as IT has been a challenge.

READ

Can Predictive Analytics Help Improve Your Instagram Strategy?

Initially IT was skeptical about the approach and was not all that supportive – the business had to get past a “this sounds like star wars” mindset. The business needed their support but IT was getting cut back and so appropriate teams were not always available. Over time the business has reduced its demands on IT as claims personnel have become able to add rules. With weekly changes to the rules and models this ability to make their own changes has been critical to delivering the agility and responsiveness they needed.

The different decisions are each handled in slightly different ways. In the Fraud Investigation Unit the analytic models and rules are used to create a credit-score like ranking. The decision is to determine the fraud risk score for a particular claim. This score is then used by the team to rank order claims so that those with the highest risk of fraud are investigated first. The decision as to whether or not a claim is suitable for fast tracking is similar, with the score being used to send the most appropriate claims to fast track. The subrogation decision is a little different with the rules and analytics being used to identify those claims where no subrogation has been identified but where it is considered likely.

Infinity is building a new claims system with some self-service options and an ability to support mobile devices and automated claims payment. This new system is being developed with hooks to the analytic decision making process to allow for no-handle claims.

Infinity found that board members can be hard to convince when it comes to claims as numbers for things like fraud are often imprecise. Subrogation was easier to “sell” in part because the numbers are easy to see – claims handed off to others to pay are very concrete. Successful adoption requires finding the pain points of executives and showing how the technology, the approach, can address those pains.