Well, thanks for joining me for our journey through the world of predictive analytics and I know I can't do justice to the whole world of predictive analytics. It's a big world, and it's changing over time. But I just want to give you some of the highlights, some of the words, some of the concepts, some of the patterns. If you think about it, I've been doing a lot of this during our time together. Talking about the past and how it's going to give us some insight about the future. Actually if you think about it right now, we're exactly halfway through our customer analytics course. So, I think it's a good time to look back at everything that we've covered so far and give you a little bit of a roadmap of what's to follow. So, what have we done? Well, my colleague, Raghu Iyengar, has spent a whole lot of time talking about data. All the different kinds of data that's available, the kinds of descriptive analysis that you can do with that data, the kinds of insights that you can draw immediately from the data basically as soon as you get it in, and that's great. For a lot of businesses, that's going to address a lot of the critical questions that they have and in many cases you can stop right there. But in many cases, the questions that you're asking, the decisions that you need to make, are going to be about the future. That's where predictive analytics comes in, that's where models come in. So what do we mean by a model? You know these days, we throw that word around a lot, we don't even think about what it literally means. It's just something we use to do stuff with data and maybe make some predictions. But let's step back and think about what model means. Think about a model airplane or a model ship, what are you trying to do when you're building a model? Well, you're looking at something that's very real, very complex, something that you can't possibly capture every aspect of it, but you just want to capture the key aspects of it. So that when we look at this thing, it's doing a pretty good job of representing something that is really complex. That's why we're building a model. Again, it's not just a means towards an end, it's often an end unto itself. It's trying to capture the reality. You know what? Customers making decisions over time, firms interacting with them, customers interacting with each other, is a very very complex phenomenon. So, one of the reasons why we build the models is to try to really do justice to that behavior. We don't necessarily want to capture every nuance of it, we want to capture the important stuff. Because once we capture the important stuff, that's going to give us that confidence to be able to make statements about what we think will happen in the next period or beyond. That's been the focus of what this particular session has been. So, try to lay out this dichotomy between models, regression models, data mining, all the things that Raghu spoke about that are really effective. If we want to make statements about who's going to do what, in the next period. Again, for many many decisions, that's all you need. Those kinds of models, that kind of thinking, has really dominated marketing analytics for a long time and I think will continue to do so. But today, as we get richer data, greater computational capabilities, and more importantly, an imperative for management that we need to look beyond just the next period, we need to start thinking about different kinds of models. Models that are better suited for the long run. Maybe we're even sacrificing something in the short run as we build these things, but we want to really capture what's worth capturing. I spent a lot of time talking about it with you, and again I tried to downplay the models. I'm proud of them, happy to give some metaphor to help you understand what these models are all about. But it's really the insights that arise from them more than the models themselves. Now, sometimes when I talk about models and I talk about predictions, and I'll show some of the kinds of pictures that I've shown you here and I say, "Hey, look how good these forecasts are." A lot of people will look at them and say, "Yeah, they're nice. But, so what?" I hear that question and it's deflating. It's like, "Come on, isn't the model good enough for you?" The real answer is, in many cases, no it's not. We need to understand what decisions to make on the basis of those predictions. I haven't really told you anything about that. I just told you about what the patterns are likely to look like if people keep flipping their coins and so on, what the future might look like. But as managers, it's our job to change the future or at least to leverage what we understand and project about the future, okay? That's where we're going to start to move from predictive analytics to prescriptive analytics. Having some idea of what the true underlying process is and what the future might look like. How do we layer on top of that some optimization. So we start to understand which message do we send to which customer at which time? Lots and lots of other prescriptive questions that managers are asking and academics like myself or more importantly, my colleague Ron Berman, are spending their lives trying to answer. After Ron is done talking about prescriptive analytics, then we'll bring in my colleague, Eric Bradlow, to tie a ribbon around the whole thing by giving a bunch of nice case studies and examples of descriptive, predictive, and prescriptive in action.
