So what else can we do with the conjoint analysis? Well, we can do conjoint on different pre-specified groups of people. Now if you want to do that, I need to remind you, you have to ask questions of people upfront in the conjoint analysis about themselves, in order to use those variables for segmentation later. If you get to the end of the conjoint analysis without asking these questions, then it's very difficult to do segment level analysis. We can also get individual level estimates. This is a relatively new phenomena, some improvements and algorithms allow us to do that. But really what it means at the most basic level, is that Ron Wilcox can go through a conjoint analysis, answer some questions, and the software can provide my utilities for the attributes. And you can go through a conjoint analysis, and it'll provide individual level estimates for you. And sometimes that proves to be very valuable. And finally, we can do something Propensity Modeling, which is trying to predict what probability that someone will buy one product verses another product. Now I want to show you what individual-level output looks like. This is some real individual-level output from my research. So you have respondents on your left-and side, just 1 through 14. And I'm showing you just a small sub-sample of the overall data that I had. And then over here, BMW Sedan, that's the utility for BMW Sedan. BMW Coupe, Ford Focus Sedan, you can see that this respondent number 1 really liked BMWs and didn't like Fords very well. But you can see that the utilities are all over the place, depending on which particular respondent is taking the conjoint analysis. You can use this information to look at attribute importances at the individual level as well. So the system will calculate that for you. And sometimes this provides very interesting insight. If you go down to respondent 13, you can see they put a 30% decision weight on price. Whereas if you look at respondent 11, they have more than double the decision weight than respondent 13 just on price. So price is a lot more important to them. Now what was I doing with this kind of individual-level analysis? Well, I'm just going to give you a brief overview of some research that I was doing. I was looking at student loan debt, to see if student loan debt increased or decreased people's price sensitivity for a major purchase like an automobile. Now, in a rational world, if you take on a lot of debt, what you might expect is that you would have increased price sensitivity. You have a lot of debt, therefore you have to watch the amount of money that you're spending on a car. In reality, what I found was just the opposite. If someone takes on a lot of debt, even controlling for other factors, this causes them to be less price sensitive with major purchases. There's something in the psychology literature, literally called the what the hell effect. And what it means in this context is, once I'm in so much debt, why don't I just go ahead and spend more money. Yeah, I know the BMW is only $10,000 more than the Toyota, but that's not a lot of money. I already have $100,000 worth of student loan debt. So that's what I was doing with my research, and I also showed that if I took people's student loan debt and showed them the monthly payments that were implied by that debt, that I reversed the effect. That the increases in debt made people pay more attention to their car purchases, and made them be more price sensitive. Now what else can we do? Well, those are pre-specified groups of people, the individual-level estimation, and now what I want to move on to is propensity modeling.