Subscribe to DSC Newsletter

I recently have been working for a client who develops predictive models for the property and casualty insurance industry. Is this really the next "frontier" of analytics? You know, like database marketing was 10 years ago?


There seems to be big demand for the quantitative master's or PhD level person who wants to do pricing, underwriting and claims models.

Would this be attractive to a good statistician? Is "an entrepreneurial statistician" an oxymoron? I appreciate everyone's thoughts

Tags: insurance, modeling, statstician

Views: 405

Reply to This

Replies to This Discussion

Without entrepreneurial statisticians, we'd still be counting out means on our abacuses. (No standard deviations.)
From the below blogs, there seems to be all this hype about statisticians and math and that perhaps actuaries do have the right skillsets. Make no mistake, there is much more to this issue than just math and stats.
I do believe that like in all data mining and analytics projects, it is less about the math(albeit it is important) and more about the data and how to work with it. This is the value-add and the reason why firms such as ourselves are engaged in this area in Canada. If you do not understand data and the proper discipline in dealing with it and yes there is a science to it, then you are setting yourself up for disaster.

I know many of you may disagree with me and yes I do have a love of math and science and how it can be applied to business problems, but the key to success is understanding the data.
Simple or even rudimentary (but robust) statistical model with deep domain knowledge does much better than very sophisticated predictive model combined with poor data or lack of domain expertise.
yes
Just to be clear i was agreeing wholeheartedly with Richard Boire's comment - wholeheartedly, let me give an example with which I am struggling right now. The governments all over the world are pumping huge (the word does not depict the size) amounts of money into the financial system right now and long term interest rates are rising precipitously, the logical conclusion is that money supply is ballooning and yet official money supply data both in the US and Europe is so noisey that one cannot clearly see M3 growing not as clearly as one can see in rates - thus one has to be apply a wee bit of mathematical trickery (statistical analysis) to these vectors to get the real picture, in effect turn down the noise control and extrapolate the trend, that is what we do, put simply, is it not?
J. Matthews--Would you be open to learning about an innovative way to give your team more access to your data and bring back your queries in seconds? I work for a company in Netezza that is the leader in this space.
I think it would qualify as the next frontier of analytics because it requires a rare combination of domain expertise, large scale computation (cloud computing), AI and statistical inference.
I don't know much about the underwriting practices in the insurance industry (my background is in credit cards and home loans) but from your question, it appears that the insurance industry is shifting towards the credit card model of underwriting. During the go-go years between 1998 to 2007, the credit card companies sent billions of pieces of applications and had develop extremely sophisticated predictive models using unimaginable amounts of data to discover the key characteristics that would lead to higher response rates. Then when the applications flooded in, they had to develop robust and extremely sophisticated data-based automated underwriting decisioning systems that could approve or decline an application within seconds while still maintaining the future credit loss rates at the same levels. There are sister systems for managing the credit lines for existing cardholders. If you recently got your card cancelled or credit line decreased, you can thank these automated decisioning systems that run the latest and greatest predictive models based on the more recent economic conditions.

It appears that the insurance industry is starting to adopt some of the credit card practices with regards to processing applications and claims. That would presumably include developing predictive models for forecasting losses based on historical insurance data and underwriting claims/applications using these models in an automated system. So the keys to their efforts would be as follows:

1) Developing predictive modeling talent
2) Developing easy-to-access historical data for predictive modeling
3) Automated underwriting - developing insurance decisioning systems that can make application and claim decisions on the fly with little or no human intervention.

#2 is probably going to be the most difficult as you need numerous vintages of historical data to develop and validate the models. As I understand from others' commentaries, such historical data are not available easily or are in paper form.

In any case, it would be interesting to see the human underwriters being replaced by the automated decisioning systems and the attendant productivity gains that would come from that shift.
I have really been enjoying everyone's discussion and perspectives on the topic. Meanwhile, the insurance predictive modeling part of my executive search business continues to expand. It seems that not only P&C insurance companies, but also large consulting firms (and small consulting firms) as well as other analytics solution providers are getting into the act.

As I learn more about the type of deep domain expertise that is necessary for this type of position, for example -analytical product development in the insurance solutions space - I wonder what would attract the top talent?

I often get asked this by clients - "how can we stand out from the crowd" of employers looking for this type of talent?

Your thoughts are much appreciated - and Happy New Year to everyone!
Marybeth
I think that in general data will never be of good quality. There is a task of professional to tackle the problem.

Of course, some lack of good data can be replaced by deep domain expertise. However, I feel the time of broad holistic views is coming, and therefore the demand for domain expert talent will be limited. Instead, talent able to deeply understand the Everything will be more needed.

RSS

On Data Science Central

© 2019   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service