The way I always like to think about it. This is one of my favorite talks that I've given maybe once or twice in the past is something I called The Golden Age of Marketing. I remember 20 years ago, again this is my pre-Dupont days. This is when I was a college student working in a marketing firm doing telemarketing. So we were calling people on the phone, and we're record the conversations. And we'd measure whether they'd respond, or even pick up the phone. And we'd collect this database of what we knew about people. And we thought man, oh man. Imagine how far technology's come. Now we get to know, we get to record the people. We could code what they've said. We thought we had great data. Well, if you went to a firm today and said, hey, I've got good news for you, this is what you know about your customers. If you remember stage one of what I talked about, phase one is the data. If you took the data from 20 years ago and that was your data today, you should be fired. It's not the best data today, and again, when I wanna speak about the next five or so minutes, is that there's always better data out there, but don't get paralyzed by this idea of waiting for the perfect data. Because I always like to say the perfect data is never coming. You have to analyse the data you have, but don't throw away data you could be collecting. So again, technology is what's driven our ability to collect data better and better. So again, just as a brief outline I'm gonna first give you kind of a historical data in marketing. So what are the kinds, thinking back to my five step program, what are the kinds of datasets that are available in marketing today? Second, what are the most emerging datasets in marketing today. And so, many of these will shock you cuz I think if I told you Google has amazing data you'd be like, of course. Amazon, of course. There are lots of companies you haven't thought about that have amazing data today. Then I wanna talk to you about some really futuristic data, like mobile data, path data, like where you are. So imagine a company knows not just what you've done but where you are physically standing when you did it. Another one that I've spent a lot of time working on recently is eye tracking data. So imagine a firm can know not only what you bought or what products you didn't buy, but what you looked at and didn't look at, and we'll be talking about that in detail. Next, predictive analytics. And this is kind of the third, and the fourth parts, and the fifth parts of my framework where I'm gonna talk to you about real firms that are doing real projects and real predictive problems using analytics. And for the part that might shock and amaze you is every example I'm gonna show you at the end today is not only real, but it's being done today. These aren't like 20/50 they're gonna launch this, they're actually using these today. So if you think about the outline that I've just presented to you, the first two parts talk about kind of, if you'd like, step one of what I talked about in my introduction, which is what are the kinds of datasets that are out there in marketing today? So hopefully for many of you watching and listening to this, you can think, hey, I could collect that kind of data. Why don't I reap the benefits of this great data? Kinda part three will be talking about what are kind of, if you'd like, futuristic datasets and how those could be used. And part four, will be the application of these by large firms today. So if we were sitting, I'll just briefly give you a history lesson, you probably didn't think you're gonna sign up for a Coursera course from the Wharton School on Marketing and Analytics and get a history lesson. But I'm gonna give you a brief history lesson of marketing in the 1950s and in the 1950s, if you like, the golden age of data in the 1950s was, I would know what was happening at each store. Now many of you may be sitting in jobs today, where store level data is the kind of data you have, in other words, that's the level of granularity you have. Now, let me say the good news. The good news is many firms make decisions at the store level. What prices should I charge in this store, what should I put on the end cap in this store? What assortments should I have in this store? And if that's the level of decision making you have, then store level data, even though it's 60, 70 years old could be fine for the decisions you have, but it doesn't allow you to do customer analytics. Again, in the 1950s, the kind of business questions you can answer was how do store level prices relate to sales? So imagine building a regression model which was talked about in the earlier modules here on analytics, where if you have sales as your outcome variable and prices as your input variables. Imagine building a regression trying to predict sales from prices. Now if you have data at the store level on prices, you can start thinking about optimizing prices. What's the effectiveness of coupons? So you can imagine knowing in each store how many coupons are redeemed at a given product category. And trying to understand how much is bought. You could try to understand regional or geographic differences. You could also understand the effect of in-store promotion, but understand, these are all relevant questions in 2015 and beyond. But in the 1950s, cuz the data was at the store level, you couldn't do customer level analytics, you could do store level analytics. And for many of you, this still might be the relevant problem. My interest and my suggestion to you is, push your firm towards collecting greater granular data, and move from store level analytics to customer analytics. And again, this still leaves a lot of money on the table. And the reason again is, making decisions at the customer level versus the store level, there's a big difference in the way that you can monetize your data that way. In the 1960s and 70s, kind of I say direct mail ruled the earth. And I say this with the greatest of fondness because most marketing analytics, business analytics today, actually comes out of the direct mail literature. So when I was a child growing up in the 70s, you would get sent your home, I remember my parents would be sent these types of mailers and flyers. And of course, the company knew which mailers and flyers they sent you. As a matter of fact, they could vary it household by household depending on what you bought in the past. If you'd like, direct mailing in the 60s and 70s was the beginning of customer, or if you'd like household level analytics. They knew what you bought. They could control which flyer, or which brochure, or catalog to send you. So they could do target marketing if you'd like. They'd know your reaction to it. They could run an experiment, which was talked about in the previous modules, when they could randomly assign different households to receive different things. And they could see the impact. Now of course, the downside of it is they know what they sent you for this type of advertising, but not other advertising. And they don't know what you bought outside of their firm. So kind of the new answerable business questions in the 60s and 70s was, now you could relate prices, individual-level prices to category and shopping behavior, but again all they would know is the prices in their catalog. They could to try to measure the impact of frequency and timing of catalogs on purchase behavior. So think about that. That was done in the 60s and 70s, think about how relevant that still is today. A firm is sending out emails. What frequency should they send them out with? What timing? Like should they send them one day apart, three days apart, five days apart? All of this comes from the direct mail literature. Product assortment, maybe you send the Bradle household, the thin catalog and you maybe send his neighbor the thick catalog. So how does the assortment you send people affect what they buy? And then finally, the type of advertising, one of the beautiful things about customer or household level analytics, is you could test different advertising messages. You could even test, should you have a red background versus a blue background, should you have 16 point font versus a 10 point font. One of my greatest quotes that I've ever heard of all times was from one of my colleagues in the marketing department here Jerry Wind. He said, only run experiments if you want to know the answer. And so, to me direct mail from the 60s and 70s laid out the five point plan I talked about. They knew the kind of business questions they wanted to answer. They knew that they could measure things at the individual household level, at least what they sent to you. They knew they could vary things. They knew they could build models. They knew they could target people, and they could make make business decisions on the basis of it. Now, of course, it leaves still a lot of money on the table because they would know what you bought in their catalog, but they wouldn't know what you bought at your grocery store, what you bought at a big department store. All they would know was their vertical, what did you buy at their. So they didn't even know what you bought at your competitors. So things like dollar wallet share, which are so important in customer lifetime value, they wouldn't know. Now let's think about the 1980s. This was kind of what most people consider modern marketing science, and the reason is because store scanners came. So we're all used to that today, but I remember a day when store scanners didn't exist. So, you didn't go to a checkout counter at a supermarket or store and everything was scanned. That wasn't the way it used to be. And so, except for what's called diary panel data, you really didn't know what individuals households bought. Now of course, you have the most granular data. And again, it's industry vertical agnostic. Clothing stores have this. Sporting goods stores have this. Supermarkets have this. Everybody can track individual level data. Now the beauty also is, let me relate this to two other very important topics, which is number one, loyalty programs. So, customers always think that loyalty programs benefit them, and they do. But remember, the minute you show your loyalty card at a supermarket, they know it's you, or the minute you pay with your credit card, they know it's you. So now, they can create what's called a panel dataset, where they're able to measure you over time, which means they can do target marketing for you over, towards you over time. So one of the things I always talk about is loyalty programs, credit card data, etc. While you see it as a customer benefit, and it definitely is, there's no question it benefits the firm cuz now they can do customer analytics towards you. Because if you paid with cash every time you went to the supermarket. How would they know it's you. They don't know it's you. You're paying with cash. They know what somebody bought but they don't know it's you. The key to customer analytics is linking behaviors over time. The scanners in the 1980s allowed the beginning of this. So again, the 1980s is the beginning of the modern age of marketing because number one, you could distribute discounts at checkout. You could track customers over time. You could give people coupons, as a matter of fact, that's still done today. A very well-known firm, Catalina, is kind of the master in giving out coupons. That I buy Pampers diapers, they give me a discount for Huggies, etc. And again, it gives me greater knowledge of the in-store experience. Now again, this is great data but it's still not the golden age of data. So imagine you had that data today, you'd be missing one big piece and that's what I wanna talk about for the 90s and 2000s and that's lead to the greatest insight and that's really the internet. So the internet really has changed the face of marketing but maybe not for the reason you think. The reason you might be thinking is well. Selling product. The internet, selling product over the web has changed the face of marketing. And the answer is, you're right, it is an alternative distribution channel, but actually, that's not the reason. Many people may not know this but, let's take a pause here, I'll let you think about this to yourself for two or three seconds. What fraction of goods in the United States do you think are sold over the internet? So, this is me sitting in front here, pausing for two or three seconds. Suppose I told you that it just hit 6%. Most people think that number's much larger. So, from a customer analytics and a marketing perspective, the internet isn't so much a second channel to sell stuff. It's a way to measure what you're looking at. To deliver you ads in a very contextualized way. So I know you're on ford.com so I can send you a car ad. I know you're on a particular website, maybe a vacation site, so now I can now only show you ads on that site, but I can re-target you later on with, hey, I saw you looked at vacations in France and you're actually, it's ten minutes later or even two days later and I'm still targeting you with ads for vacations in France. The internet has allowed firms to do the ultimate in customer analytics. I always like to think of the internet as being kind of, now that you have internet data as well as purchase data, if you have those fused together, I almost think of this as an emergency room doctor which is you go into an emergency room, you're bleeding. What does the doctor do? He wants to know why you're bleeding. Well my view is, if you're a firm and you're manufacturing a product, and you wanna know what's driving your sales. I think you wanna know not only what people are buying, and that scanner data from the 80s could tell you, but you wanna know what products people are considering and ones that they haven't chosen to buy. So internet fused together with purchase data. If you'd like, take your weblog data and fuse it together with your supermarket and store scanner data. Now you have a very good picture of advertising and of course, if you can add the third piece which is TV, which is starting to come. Now all of the sudden you have an integrated view and the data that you really want on costumers to do analytics towards them. So again, from internet data you can track which pages they've gone to. Probably, the most important part is which products they consider, cuz one of the ways I always like to think about it is, imagine two scenarios. One scenario is you have low sales because people considered your product and don't like it. Another one is you have low sales because people didn't even consider your product. I'd rather have the latter than the former. The former means everyone knows about your product, they've considered it, but they don't like it. The latter means maybe you just need to spend more money on advertising to generate awareness without the data on what products people are considering, you don't know it. Targeting ads, that's kind of the big part. As a matter of fact, it's gone through its double hunt cycle. Back in the late 90s, early 2000s, banner ads were kind of the rave and then in the mid 2000s they died off cuz people said, well, they don't lead directly to click but now they're kind of back cuz people realize they generate awareness and they generate brand equity. Next, the internet allows you to link the past experience so you can track people through their IP address, through what their called cookies, their drop loyalty programs, and the last part, and this is the part I wanna emphasize to you cuz part of this module that I've been talking about is getting the right data. It's crucial that you find a way to link your online and offline data together because a lot of people now do what's called show-rooming. They go in the store, the physical store, they look, and then buy online. A lot of people look online and then buy offline. It's really important that you link those two together. Loyalty programs are one way to do it. There's ways to do it by giving people discounts and coupons, by recognizing themselves when they go in the store. You gotta link the loyalty card data to the online data. Many firms lose the ability to do modern marketing customer analytics because they don't properly link their online and offline data together. And if you'd like the explosion of new data is here. So if you look at all the different names, Google, Facebook, Twitter, Yahoo, Tumblr, Flicker, etc. It's probably as no surprise. That all of these companies have great data that firms wanna know about, but how about this set of firms? So for example, Expedia, what data does Expedia have? Well, Expedia has every time someone goes onto their site. And of course, notice whenever you want Expedia they either make you log in or they say sign in as a new customer cuz they wanna track you over time. Another example given here is Sirius XM Radio, the satellite radio company. Of course, they know your listening behavior. The wanna know, I wonder how people who listen to certain types of music if they're more likely to cancel their Sirius XM accounts. Imagine Hertz, the car rental company, one of the questions their interested is what car are you gonna rent next? Can they upgrade you? What actually is common to all of these companies, and you probably aren't thinking about all these companies. All of these companies here make profits one customer at a time. They all have individual level data that they use to monetize. And they're all applying customer analytics methods to try to make more money, and to try to optimize their marketing against it. So
