How GOOD Is Your Data?

Your reinsurers want to know.   And so does your board of directors, if they were paying attention to John DeMartini’s excellent presentation at last week’s NAMIC Convention.

Mr. DeMartini, Executive Vice President of JLT Re, didn’t ask how MUCH data you have, or how MANY records in your dataset.  More is always better than less, but there’s only so much an insurer can do about sample size.  He didn’t ask how smart your analysts are, or how many variables are in your GLM, or how colorful your dashboard reports are.

And he didn’t ask because he is curious.  Or in love with technology.  Or stymied by Analysis Paralysis.  He showed, with a few quick examples, how accurate data improves an insurer’s understanding of loss costs, leading to more adequate pricing, sharper underwriting and more confident cat loss projections.

Good data leads to better decisions which create more net income. 

So, how good IS your data?

Don’t ask your IT department.  They don’t know.  They can tell you how well normalized your database is, and how much data is stored in operational systems rather than archives, and what the database architecture is, and how much storage space it consumes, and how long the update jobs run.

But data quality is not an IT function.  If you really want to know how good your data is, ask your front line underwriters and claims adjusters.  They are the ones acquiring the company’s data.  They can tell you which data elements are self-reported by insureds and agents, and which ones are validated by an objective source.  They can tell you which data elements are entered via keyboard (unstructured) vs selected from a menu (structured).  They can tell you which data elements magically appear on a screen without any human input (automated third-party data).

Data quality is determined more by your processes than your systems.  It’s hard to do right, but it’s the foundation for your understanding of cost and rate adequacy.  Every analyst should have a working knowledge of the company’s data sources, and at least a rough awareness of the quality of each data element.  Every company leader should be asking about the business processes that lead to data acquisition.  And every board member should be asking if the house is built on rock or sand.

How good is your data?  If you’re not sure, or you don’t like the answer, then call me and we’ll work on it together.

footnote:  National Association of Mutual Insurance Companies Annual Convention, September 26, 2017. “Questions All Directors Should be Asking – Part 2, Technology, Reinsurance & Cat Modeling/Analytics” John DeMartini, Executive Vice President, JLT Re.