“If you model the risk and revenue of applicants, the type of browser shows up as a significant variable.”
So this is a new one. Not sure if you’ve seen it before. I just put my order in for my Nissan Leaf and received the total price for it so decided it’d be a good idea to take a look and see what auto loan rates were. I checked my credit union first and they currently are offering 3.99%. Not bad, but about a week ago Capital One had sent me an email advertising a 3.10% rate. I went to check the website using my default browser (Firefox 4 Beta 6) and noticed it was at 3.5%.
I figured it had just gone up since I received the email. I tried to use their little payment calculator but the flash based widget wouldn’t work properly in the Firefox Beta so I loaded up Safari to try and funny enough the rate offered was 2.7%. I checked in Chrome and Opera to see if it was maybe just something wrong with the Firefox beta and Chrome’s rate was 2.3% while Opera’s was 3.1%.
I’m not sure why Capital One would choose to offer Chrome users a lower rate than Firefox users, but it’s interesting nonetheless.
I read Supercrunchers a couple years ago, and CapitalOne figured prominently in its exploration of how people use large datasets to make better decisions. A couple years ago, someone told me that CapOne had ~70 terabytes of data, which is a ton of alphanumerical (i.e., non-image) data.
I find this really interesting, and not at all bothersome. Like CapOne, I’m a big believer in using data to improve your performance. And if they have modeling that supports “browser” as a significant predictor variable, good on them. If they’re wrong, and they find that as the n grows browser choice becomes a less stable variable, they’re the ones paying the price (through loan defaults) and they add more data to their pile. Great stuff.
Related: I have been using Chrome exclusively for non-work browsing since late 2009. Unfortunately, I still use IE quite a bit at work because we are a SharePoint/Cognos shop.