Pulleyblank: If you think about what IBM's been doing for a long time, we've been collecting and managing digital information ... and we're terrific at that.
The problem we're really trying to address now is how do you take the massive amounts of this information and give our clients and customers the ability to make better business decisions using that information.
So, the challenge of business analytics and optimization is how do we go from this massive amount of data, analyze it, process it, so people can really use it to make those better decisions.
developerWorks: Now when they start using it, what does it do for them? What kind of things happen?
Pulleyblank: Well for example, one thing we do now with it and which will continue to grow is ... fraud and abuse detection in insurance claims. Now, it's the case that if you're an insurance company and you're getting a lot of requests for payment, most of them are probably fine, there are some that you probably shouldn't be paying. It may be because somebody is submitting fraudulent claims; it may even be because a physician is not aware that a fine new test that he's using made one of the other tests that he's always used unnecessary. So he should be dropping that out.
What we provide is a solution which will analyze the historical payments, will look at the things which are "different" from what the normal behavior is, flag them, and these are the ones that they can investigate and decide whether they should pay them, should they not pay them.
developerWorks: So, just infinite variations upon that?
Pulleyblank: Huge variations. Well I'll give you another example of that. If you start with that, we've done ... we've taken the same idea and done it with the State of New York on their income tax collection. And the question they have is "Who are the people who are underdeclaring income on their income tax?".
By the way, when I describe this to my daughter, who is living in New York, she, uh, thought it was a very interesting application and wanted to know how it was done.
But it's the same idea. It says "What's normal behavior?," "What's abnormal behavior?," and let's look into these abnormal cases and see. They may be alright, but they may not be. And if we can find it and target those people, that's got a lot of value.
It covers almost every area you can imagine. I mentioned the insurance claims; we've done a project with the Metro system of Los Angeles, LA Metro. Now, this was on inventory optimization. Now if you think for a moment, "What does a metropolitan transit association do with inventory and what's the issue on that?" ...
Pulleyblank: Well the answer is it's spare parts for the cars and the trains; because when something breaks down, if you cannot repair it and get it going again, you're going to be running without the necessary equipment.
developerWorks: You need it fast.
Pulleyblank: You need it fast and you need the right stuff and you probably've got a much bigger inventory than you need because large parts of it are old pieces that you never need or parts that never fail. So what we were able to do was do an analysis of the entire spare parts inventory, say "This is how much you should keep of each particular one." By doing it you can reduce the cost of your inventory and at the same time, have the right parts when you need them.