Data Mining and Warranty Fraud
I have problems keeping my glasses clean - it is the oddest thing - dust just seems to settle right on the lens. It seems that no matter how I clean the lenses, whatever I am doing is causing small cracks in the lens where they meet the frame. So, today I went to the place I originally got the glasses made. Well, I got the manager and she went over my records - “I see we have replaced your lens before.” She then gave me a little lecture on the proper way to clean the lenses (which - I was actually doing correctly) - but since they were under warranty - she would have them replaced.
Now, in theory she might have refused to replace the lenses after seeing a pattern in my customer records that showed some potential fraudulent activity. Ultimately, she just decided that I was a goof and just didn’t know how to clean them correctly. However, warranty fraud is a big deal. Think about car or motorcycle dealers - they often do work covered by manufacturer warranty. Once the work is done, then they bill the manufacturer. Let’s say you own a Honda dealership and you do a lot of warranty work. One day you accidentally bill Honda for an extra part not actually used in the warranty work, but Honda doesn’t notice and pays you for it. That is when the evil thought enters your mind - maybe a couple of parts here and there - with an excellent understanding of Honda motorcycle specs - you can put together plausible warranty claims.
That is warranty fraud. Claiming to do work not done or using a part that was not used. The problem is - for the manufacturer, it is especially difficult to police the dealers - and potentially the cost of auditing far out weights the potential fraud that may be found. It is always a trade-off. This is where data mining can really be an excellent way to help leverage manufacturers’ limited resources in tracking down fraudulent claims. Using data mining, high potential target dealers can be reviewed by auditors - rather than doing random audits. Random audits certainly have a price in dollars, but more importantly they have a negative impact on the relationship between dealers and manufacturers. Data mining can help reduce the “randomness” of audits and hopefully more frequently target the dealers that are acting fraudulently.
