Data Quality

At all times during the negotiation for insurance cover, risk information is imperative, but certainly when there is a rush, as referred to above, when brokers and underwriters are being torn in many directions, the quality of information can make or break a deal. What follows is a short discussion on this subject.

With the correct tools, an instruction manual and enough time, most people could manage to change the oil in their car. But most of us choose not to. Faced with a portfolio of property locations suspected of containing erroneous information, it is tempting to believe that finding and fixing the problems is a trivial exercise, and possibly not even necessary. But data quality, once considered unimportant, has now become a key driver in pricing risk and increasingly is seen by rating agencies as a proxy for the effectiveness of management controls (if it can’t be measured, it can’t be managed). Improving data quality has a direct impact on catastrophe model loss output, and can impact insurance and reinsurance costs by many millions of dollars.

Analysts working at insurance and reinsurance companies are smart.  No doubt that with enough time, they could find many of the instances of wrong data. Key in a few commands to Microsoft Access and it is fairly easy to find the gaps in a portfolio. But what should be done about the missing data? Ignore it? Guess? Go back to the original source? The latter is rarely an option; guessing can be worse than ignoring the problem, but to ignore it means underestimating the loss potential. A search on Google may reveal a photograph of the building in question and support an educated guess at what it is made of, where it is and what it is worth, but is hardly practical for more than one or two buildings.

And what about data which is complete, but wrong?  It doesn’t take a degree in civil engineering to know that a 73-storey hotel in downtown Atlanta recorded as “masonry construction” has been wrongly coded.  But the average analyst looks at tens of thousands of locations a week, and is under a tight deadline to provide modelled loss assessments. Trawling through a portfolio searching for anomalies is a full time and onerous job, and many instances of wrong coding will still be missed (“steel frame” construction” may seem similar to “light metal” but is far less vulnerable to hurricane damage).

The human brain is excellent at noticing discrepancies in visual patterns, and in using past experience to apply heuristics, or rules of thumb, to know where to look for problems. We can sense when something doesn’t feel right, without always knowing exactly why.  But we struggle with data when presented as raw numbers, particularly on the scale routinely exchanged between insurers, brokers and reinsurers today. Computers, on the other hand, are very good at processing large amounts of data, but need to be trained to look for specific errors or potential.  Properly designed, automated analyses using finely crafted heuristics can point out suspicious information which needs further investigation.

There is still the challenge of how to replace information which appears wrong or is missing. What is needed is a reliable, independent source of property information which can be accessed automatically whenever suspicious building data is identified in order to provide credible alternatives to occupancy type, construction, height or value.

A consistent and credible index of data quality which can be split into its parts enables an analyst to understand the worst problems which need to be solved in order to improve the quality of a portfolio. For example, knowing the exact street address is crucial for a building located on the coast of Florida but is of limited value for a similar building in North Dakota. Such an index also helps in the communication of data quality between different parties.

Computing analytics will never - and should never - totally replace human intuition. Used appropriately, they can ensure resources are leveraged much more effectively on higher value activities.

In preparing this SPOTLIGHT article we have received invaluable assistance from Risk Management Solutions – RMS – in London (www.rms.com) for which we are grateful.

If you would like Worldwide Risk Solutions to conduct an economic, business and insurance survey of any international markets please contact us – Details below.

Worldwide Risk Solutions has access to a wide client base of internationally oriented organisations. Why not utilise this knowledge and experience? We can conduct a swift appraisal of your global activities or answer any questions you may have about international developments. Call +44 (0)1444 450 919 or send us an e-mail and we will respond immediately. And should you be passing through London, please let us know.

In our next Spotlight article we will be looking at the world of valuation, valuation-related advisory services for business, financial, legal or tax purposes as well as valuations for solvency or fairness opinions, financial or tax reporting and compliance. In light of some of the notes in our Bazaar, this will make interesting reading. For more information about any of the items discussed in the current or previous issues of WoRdS, please see our Contact Details:

George Worsley, Director
Worldwide Risk Solutions
Telephone +44 (0)1444 450 919 Mobile: +44 (0)7968 191 511
E-mail info@worldwiderisksolutions.com