“What If” versus “What Is"

Insurance carriers are obsessed with models. Catastrophe models, predictive models, statistical models – they are all the rage. Tyra Banks would be proud. Models (of the non-runway variety) are useful tools to answer questions like “what could happen” and “what might be”.  But these models are most useful when trying to speculate on the future. We should have better ways to evaluate the present and the past. 

I recently read the article “Modelers’ Sandy Estimates Could Be Misjudging Insured Losses[1]  and to tell you the truth, I just didn’t get it. Didn’t Sandy already happen?  Isn’t it in the past?  Why do I need a model to estimate Sandy damage?  Shouldn’t I be able to count and measure it directly?  I don’t see the need for a predictive model.  I see the need for a tape measure. Admittedly, models are sexy. They can offer amazing visualizations and a façade of accuracy – that’s what they do, they are simulations after all. Somewhere in all this “make believe”, it seems we’ve lost our grip on reality. 

In the field of mental health, the inability to distinguish between fantasy and reality is called a dissociative disorder.  People develop dissociative disorders in order to cope with extreme trauma - the experience in their real life.[2] Organizations develop dissociative psychoses in the same way. Have you ever worked in an insurance operation and tried to get hard data on what is actually happening? Like counting the cost of a storm, comparing the development long-tail claims from year to year, or understanding the true impact of a recently released product change? If you have, then you likely have a minor addiction to Advil if not something more potent. Getting access to real-time analytical data is a huge challenge. All the hype around “big data” feels more like a big headache than a big opportunity. It’s no wonder that we prefer hypothetical models over reality – it’s how we cope with the pain.

Here’s a suggestion: let’s invest in understanding reality as much as we invest in predicting what might be. We need better ways to understand the actual impact of storms, more current and accurate measures of performance, and better external context to make smarter decisions; the list can go on and on. Basically what I’m saying is:  We don’t need more “what if” analysis. We need more “what is” analysis.


[1] http://www.propertycasualty360.com/2012/11/13/growing-accord-modelers-sandy-estimates-could-be-m?t=erm&utm_source=PC360NewsFlash&utm_medium=eNL&utm_campaign=PC360_eNLs

[2] http://www.cnn.com/HEALTH/library/dissociative-disorders/DS00574.html