Earnix Blog > AI

Blog: Five Insights Gleaned from the Casualty Actuarial Society Ratemaking, Product, and Modeling Seminar

Jonathan Moran

16. April 2018

  • AI
  • Customer Centricity
Earnix recently attended the Casualty Actuarial Society Ratemaking, Product, and Modeling Seminar (CAS - RPM) event in Chicago, Illinois. While the conference covered many innovative topics on trends in usage-based, telematics insurance, and regulation – Earnix was there to present  specifically on customer analytics, customer lifetime value, machine learning, and analytical solutions for small insurers. The conference also coincided with the launch of the  Earnix Integrated Machine Learning™ technology.

So what did we learn? Lots of information was gathered from attendees through live polling during the presentation sessions. Session attendees were pricing and analytics professionals for large P&C and Life carriers in North America. In this blog post, I’d like to share the five insights I found most valuable. These insights include:
  1. Spending on analytics is still a priority.
  2. But data access and management is still a challenge.
  3. Insurers want to know how to apply machine learning to their business.
  4. In order to do this, they need access to best in class technology.
  5. Once they tackle these issues, creating metrics like customer lifetime value essential.
Earnix delivered a session focused on practical advice for small insurers. The presenter, Drew Lawyer, polled the audience for some interesting preliminary questions that really set the stage for the presentation. During his polling, he found that the majority of the audience consisted of P&C carriers with greater than 500m in annual premium volume. A third of the audience had 2-5 people in their pricing analytics unit, while another third had 5-10 people. The remaining third was split between less than 2 and between 10 and 20 people. His next two questions were perhaps the most telling. He asked:

“Over the next 12 months, how do you see your investment in analytics changing?”


This demonstrates the fact that spending on analytics is still a priority. While this is a great sign for companies like Earnix, Drew then asked the following question which leads us to our second insight.

“What do you believe to be your biggest challenge when it comes to enhancing your pricing analytics?”

“What do you believe to be your biggest challenge when it comes to enhancing your pricing analytics

Not surprisingly, data access and management is still a challenge. Not having enough of the correct data to fully develop analytical programs is more of an issue than analytics resources, tooling, IT barriers, or the corporate culture.

Armed with this information, Drew delivered practical advice around how to develop culture and staff, how to plan an analytics roadmap, how to perform loss cost modeling, how to source and work with external data and tools, and finally how to leverage new machine learning techniques inside of smaller organizations. At the end of the session, Marcus Deckert of Pekin Insurance took the stage and gave some more practical advice, based on his experience that smaller insurers could leverage.

Machine learning was a hot topic at the conference this year, and Earnix delivered two sessions that focused on the application of machine learning techniques inside of insurance. The audience was polled on where they though the insurance industry was heading from a trend and technology perspective.

“Where do you see the future of the insurance industry heading?”

“Where do you see the future of the insurance industry heading?”

Almost half of all respondents saw AI and Machine Learning to be top of mind for the insurance industry, and Earnix would certainly agree with this statement. The question was then asked:

What is your sentiment toward the following statement: “Why are you not getting value from your Machine Learning and Data Science?”

data sciance

This answer speaks to the fact that most insurers are just getting started with Machine Learning and want to specifically know how to apply it to their business. Based on these results, Reuven Shnaps then moved into practical applications of machine learning – including applying machine learning to the price setting framework, rank optimization, lifetime value application for underwriting, predicting product acquisition sequence over time, and product personalization.

Naturally, a machine learning library of models and algorithms is only as good as the underlying technology used to create the advanced analytic program. A question was posed to attendees, “What software do you use for data analytics and modeling?”

data analytics and modeling?”

As expected, R is preferred currently over SAS or Python for its open source availability, ease of use, and low associated costs. This speaks to the fact that access to best in class technology is essential for insurers of today.

What we saw clearly from attendees was that developing robust analytical programs for insurers both large and small is comprised of several key factors – a solid internal team and business partner willing to serve as champion, sound data and data practices, the ability and knowledge to start small and gain quick wins and having the right tools and culture in place. Once all of these items are accounted for, traditional customer behavior data attributes can be combined with machine learning based customer demand modeling for the development of a complete customer profile. With this information, advanced metrics like customer lifetime value can be calculated and used in marketing, sales, service, and retention programs. But where are insurers as far as maturity around these advanced metric calculations like customer lifetime value? The question was posed:

“For how many of the following 7 functions do you use a LTV metric? Pricing, Underwriting, Claims, Next Best Offer/Action, Retention, Marketing, and Acquisition. “

Acquisition. “

Over half of respondents (51%) are using 1 or fewer functions with a CLTV metric, with 18% of respondents measuring 4 or more functions with a CLTV metric.

We then asked, “If you currently use an LTV metric for more than one function / activity, do you use the same metric across all functions?”

For this response, 50% of respondents had more than one CLTV metric in play across differing functions. For the other half of respondents, the question was either not applicable because they weren’t calculating CLTV or they simply used one metric across all functions they were tracking. This leads to our last insight, which is calculating CLTV is essential to aid insurers in informing pricing and product decisions.

Earnix spoke to over 350 attendees at the conference, delivered three outstanding sessions, gave away lots of cool prizes, and even had some fun while doing it! If you were able to join us at the conference – thank you – and if you weren’t, please visit our website to check out all of our newest content and ideas!

Teilen article:

Jonathan Moran