Predictive Reasoning

Devastatingly smart Rafal Lukawiecki, that virtuoso of technical presentation, gave yet another brilliant presentation on how the seemingly “ivory tower” techniques of Artificial Intelligence can be used in present day applications for unusual data validation scenarios.

After using the Irish Postal service as an example of where statistically aberrant data is required to actually get a letter delivered (for some unfathomable reason), he then went on to describe the application scenario. By using the Business Intelligence tooling available for SQL Server 2008, Rafal dazzled us all with the simple way his sample data validation App predicted that the user had submitted dubious data, and importantly, without writing one single rule (or more correctly only one single rule!) reported exactly what was dubious about the submitted data.

In essence, how it works is, when a statistically significant set of valid data exists, it’s possible, using the Business Intelligence tools, to construct a pretty simple model of what can be considered correct and appropriate for submitted data. For example it’s probably not appropriate to have a female son. And what’s more, it can even tell you exactly what it thinks is inappropriate in the submitted data (which may or may not be politically correct!).

The potential benefits for insurance adjudication software immediately jumped out at me. Plan membership and care provider matching projects are huge areas of development and refinement in this business. Matching submitted data to appropriate recorded data for members and providers, with a good level of confidence, is mandatory for a cost-effective claim adjudication system.

Claim adjudication efficiency can be easily determined by a simple metric referred to as the “drop-to-pay” rate. Drop-to-pay is just the percentage of received claims that reach a finally adjudicated state, without being touched by human hands, as it were. If you have a system with high drop-to-pay rate, then you have an efficient, cheap to operate system (in theory!).

Clever use of existing historical matches using a solution similar to Rafal’s approach, could potentially yield much higher match rates than those presently obtained, thus increasing a system’s drop-to-pay rate significantly. Remember a rate change of a percentage point or two can make a very significant cost saving for the carrier.

When I queried Rafal about it after the session, after admonishing him for denigrating the wonderful Irish Postal service, he told that yes, this technology absolutely would translate into a very successful matching technique learning constantly from its successes and failures. Failures being more educational than successes as in the real world!

  1. Leave a comment

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: