Call us at 650-400-3029 (PST)

Tips for Successful Data Science Implementation in Insurance

Nancy Casbarro and Deb Zawisa of Novarico recently published a new paper on Data Science in Insurance: Expansion and Key Issues subscription required) that was summarized in this nice little article on Dig-in  3 challenges facing insurers in data science implementation. These three challenges – getting business buy in, attracting talent, and strategic alignment are exactly what we see in our work with insurers. And we think our DecisionsFirst™ approach addresses (or helps address) all three.

1 – Getting business buy-in

We begin our DecisionsFirst approach by working directly with the business SMEs to define a decision model. This logical, structured view of how they want to make an operational business decision (such as paying a claim or underwriting a policy) focuses the initiative on a business problem. Once we have this model we ask the SMEs to imagine what would help them make it better – “if only we knew XXX we would decide differently”. And these XXXs are almost always the kinds of things that analytics, machine learning or AI can identify such as customer mood, likelihood of deception, risk of waste etc. We work with the business SMEs to determine how they would change their decision-making if they knew these things and how accurate they would need these new facts to be. All this frames up analytic requirements, so the analytic team can be quickly productive. And more importantly, it ensures that any analytic the team builds can be readily adopted and deployed because the business team already know where it fits.

This approach also supports the critical test and learn skills the authors identify. The decision model can have the current way and the new way integrated into it, so new analytic approaches can be trialed and compared easily. The decision model gives a clear, unambiguous definition of the decision-making being tested.

2 – Attracting talent

One of the best ways to solve the problem of attracting talent is not to have to attract so much of it! We have found that teaching decision modeling to business analysts and data analysts makes them much more productive when working with data scientists. Getting good at understanding how decisions are made and how analytics can improve decision-making is often a first step to becoming a data scientist and allows companies to identify those with the interest and nascent talent for decision-making solutions. These folks can be encouraged to take advantage of available training and modern analytic tools and frameworks to grow their skills. It’s not a substitute for hiring experienced data scientists but it does start to grow your own.

3 – Strategic Alignment

The focus on engaging with the business experts, not just the data, right at the beginning of an analytic undertaking is critical to getting alignment. If you listen to the business before you listen to the data, you can understand their challenges, their needs, and the kinds of solutions that will help. This will make any data science work better aligned.

More than that, though, is that decision models are a great tool for operationalizing analytics. Each sub-decision in the model can be identified as needing an analytic algorithm, being something best left to people, or as something for which well defined business rules can be specified. Using a Business Rules Management System (BRMS) to automate these sub-decisions allows a significant amount of the decision-making to be automated and wraps that automation around the model to put the results to work. Using a DecisionsFirst approach and then mixing business rule and analytic technology into a decision-centric solution generates a real business return from your analytic investment.

I’ll let the authors have the last word with this great summary of the key problem in data science:

“Commitment to operationalizing these insights is the next step that can be lost in funding or prioritization battles. Even if data science initiatives are aligned with the business strategy, the only way to get real value is by operationalizing the results.”

To learn more about how insurance carriers can profit by focusing on decision-making, read this whitepaper.