Bob Long: When you buy a book on Amazon, you probably notice you’re likely to get a message suggesting some other books you might like to try too, similar to the one you just bought. You may have also had similar experiences by swiping your loyalty card at your favorite grocery store or clothing store. You’ll get in store coupons or emails from the business offering items they know you seem to like. I’m Bob Long; we welcome you to another edition of Stats and Stories, a program where we look at the statistics behind the stories and the stories behind the statistics. Our topic today is Personalizing the World, how businesses have moved from mass communication to a very highly personalized communication. To prepare you for our discussion, Stats and Stories reporter Justin Maskulinski looks at the tectonic shift in how businesses are doing marketing today.

Justin Maskulinski: There was a time in the not-too-distant-past when companies relied on mass marketing to convince you to buy their products. TV and radio commercials, newspaper or magazine ads or direct mail were considered a key part of marketing. But the Director of Interactive Media Studies at Miami University says marketing has undergone a dramatic shift. Glenn Platt says using sex and sizzle to sell products has been replaced with tons of personal data that marketers can use to reach us.

Glenn Platt: The game of marketers changes from being about convincing people that this is going to change their life or sort of drawing people toward their product and instead saying: Hey, I know something about you. I know what you like. I know what makes you happy. This is a utility proposition here. I can make your life better.

Maskulinski: Miami Marketing Professor Jim Coyle agrees. He says your personal information is valuable currency that marketers can use in a positive way.

Jim Coyle: We're all learning, right? The marketers are learning what we as consumers want, and consumers are learning how valuable that information is and what they can do with it - how they can spend it to get something of value back.

Maskulinski: Glenn Platt says companies like Kroger use Loyalty Cards to determine what you buy and offer you coupons they know you'll use. He says that's more effective than having a celebrity trying to sell you products…

Platt: And so for marketers at Kroger, Loyalty Cards create this opportunity to say, look. Here's this win-win proposition. We can find you things that are consistent with what you've bought before that are cheaper, better, faster. In exchange for that, we know you're going to come back to Kroger more often because we've customized that.

Maskulinski: Jim Coyle says Kroger and Amazon are great examples of companies that have used our personal information to offer things of value. But Coyle says it's critical for companies to find a fresh approach…

Coyle: I think the problem with the loyalty cards can be that they can get kind of old. And it's easy if you're a marketer and you're doing a great job with the loyalty program, and I'm a marketer and I see what you're doing; I can copy that, right? So there's always this need to make sure that your loyalty card goes above and beyond what the others are doing.

Maskulinski: Personalized marketing can be helpful, but some consumers are concerned about how companies share their data with others. Coyle says if you look at the bottom of the page when you're online, you'll find boxes that warn you when there are cookies on the site that will be following what you do. He thinks marketers are trying to be more open.

Coyle: I think that they need to keep it simple. They need to - when you first go to a site or when they first invite you to join a program to sort of pull out the most salient points of their privacy policy that are going to be most relevant to us as consumers and make sure we understand them and also give us the opportunity to delve further into those privacy policies if that's the sort of consumer that we are where we're really concerned about what sort of deal we're making with them.

Maskulinski: Glenn Platt envisions a day when we will have more control over marketing.

Platt: I actually think the future of this is where each of us becomes a chief marketing officer of ourselves. There will be a CMO of you, and that is you. And you manage your data, and you decide who gets your data and who gets paid for your data and how your data gets used. And if someone is making money from your data, you want to be the beneficiary of that.

Maskulinski: Glenn Platt says companies must be transparent about how they use our personal information or who they share it with. For Stats and Stories, I'm Justin Maskulinski.

Long: Joining me on Stats and Stories for our discussion of this new business model of highly personalized communication, Miami University Statistics Department Chair John Bailer and Media, Journalism and Film Chair Richard Campbell. Our special guest today is the Managing Director of Customer Knowledge at dunnhumby, Paul Hunter. He’s in that field of trying to find out, speaking of feelings, what customers feel and how they do things. And that helps dunnhumby’s clients improve their strategies. And he has over 25 years of experience in this field. Paul works with some of the world’s largest retailers and brands and Paul, I just kind of wanted to jump in right there because a lot of people might not know much about dunnhumby and exactly what you do and how marketing has really changed in the last decade or so with all the new media technology that is now available.

Paul Hunter: Well first off, thanks for having me, it’s great to be here. And definitely the technology landscape has changed, not just in the past ten years, but it’s probably really been the past twenty plus years and it really started I think, in the consumer packaged goods world with the advent of the barcode. The barcode was first introduced in 1973 and that really began the digitization of what people buy and you used to be able to get data on individual cities or individual stores in terms of what people were buying. And then around, really the 1980s or so, loyalty cards began to become mainstream and that took it down to another granular level of you could begin to pair up what people were buying to the individual household. That’s really what dunnhumby does for its clients, is look at the household level data and it’s really altered marketing. I would say over the past ten years in particular, using your timeframe, from going from marketers just doing mass advertising on TV, on radio, and they still do it, but it’s opened up a huge avenue of really talking to people one-to-one as opposed to a mass world.

Long: John Bailer, we go to you for the next question.

John Bailer: So Paul, our lives certainly have changed with the introduction of using these loyalty cards. It was something that was inconceivable for most of us even a decade ago and now it’s so routine. So one question is what happens when we do this? When we enter our loyalty card number or it’s swiped, what data are being collected from that and ultimately how might that be used?

Hunter: It’s a very good question. I’m going to actually start a little bit back because while we call them loyalty cards, they’re not necessarily a measure of loyalty. If I go to Chicago and I talk to somebody on the street, I might find that they have a Dominick’s card, a Jewel card, a Walgreens card, a Rite-Aid card, so a lot of people have different cards. And from that 1980s through today, every retailer has got these loyalty cards, but that doesn’t necessarily mean you’re loyal to them. It’s a great mechanism to pair up what people buy to the individual household and that triggers a recording in the database of the retailer that they understand that household 1, 2, 3 purchased this bottle of water, this brand of laundry detergent, this brand of toothpaste, but then the retailer has to understand, okay we have millions of these households, can we begin to figure out who’s loyal?

Bailer: So how’s loyalty defined? Do you have to operationalize that in some way?

Hunter: Absolutely. And a retailer only has the data from their own stores so they don’t have the liberty of looking at, Walgreens doesn’t have the liberty of looking at what Rite-Aid has and vice versa, so what somebody like Walgreens will do, they will look at a very common statistical technique or mathematical technique and it really began in the 1950s with direct marketing. It’s called RFM which is Recency, Frequency, and Monetary Value. So it’s an RFM segmentation that doesn’t really use fancy statistics, but it uses basic math. And it says take it at the household level, sum up all the data in the database for each individual cardholder, look at the recency, when were they last in our store, the frequency, how often do they come into our store or a certain time period maybe it’s two months or six months, and then the monetary value, how much do they spend with us. So you can begin to get a measure of loyalty by looking at that recency, frequency and monetary value of your individual households. And so for a retailer like Walgreens, they might find that while they hold data on 80 million households, really it’s 8 million that make up 80% of their sales.

Long: Quick question on that too. Let’s say I go to Kroger. I think a lot of people think, oh I swiped my Kroger card and I saved X number of dollars, the cashier usually says you saved so much money this time. But what you’re looking at are what are my habits, does that also help the stores now determine what items they may want to offer on sale to their loyalty card customers in general?

Hunter: Absolutely. Taking a retailer like Kroger, and the other thing you have to keep in mind, going back to that frequency, a grocery store is a wonderful opportunity to get more observations. So statistics, particularly when you are looking at time series analyses is we love data on an individual household that’s plentiful over time. So if you’re really loyal to a Kroger, or Safeway or Walmart, and they’re your main shop for groceries, the typical loyal shopper will be in there 70 times a year. Where maybe like a Walgreens you may only see them 10 times a year and let’s say a Macy’s or Target, you might only see 3-4 times a year. So the grocery store is an environment that is rich with observations, rich with time series analyses so you’re really limited only by what I would call your statistical imagination in terms of what you can do with the data. And so one thing that we can begin to do is paint a profile of your lifestyle. So we can pretty much tell, when you registered for the card you didn’t tell us that you have a pet, we can tell that you have a cat based on what you buy. When you registered for the card, you didn’t tell us your age, but we can pretty much approximate what age you are based on the products you buy. You might be buying diapers for children. You might be buying different types of food for a certain type of dietary needs, so there’s a lot of rich information that we can begin to distill through the behavioral data that comes through the till. And the going back to the question of loyalty, we know that you might hold four or five cards in your wallet, but the challenge now is back to the retailer in terms of if this is really a loyalty card, what are you, as an organization, going to be doing differently for your best customers? And so the question is the loyalty card is not necessarily customers being loyal to the retailer, but we would challenge that it’s what would the retailer do to be loyal to its best customers? So there comes your point on promotions or particular offers that are highly customized, highly personalized and hence I think the title of this episode Personalizing the World is, retailers have rich data, almost too much data that they can shake a stick at it, but if you use the data intelligently retailers can be very loyal and serve up highly customized promotions where if you have a cat, why don’t we give you a free bag of cat food and thank you for coming into our store. And so that’s how it’s used.

Long: You’re listening to Stats and Stories where we discuss the statistics behind the stories and the stories behind the statistics. And today we’re talking about personalizing the world, how businesses are moving from mass communication to highly personalized marketing. I’m Bob Long; our regular panelists are Media, Journalism and Film Chair Richard Campbell, Statistics Department Chair John Bailer and our special guest from dunnuhmby is Paul Hunter, the managing director of Customer Knowledge. And Richard Campbell we’ll go to you for the next question.

Richard Campbell: Paul, following up on the loyalty issue, what do you do with issues of privacy? I mean we have lots of concerns about how much companies know about individual behavior and how do you handle that sort of problem with customers because it’s probably the flip side of loyalty, concerns about these companies know way too much about me, is that a terrain you have to maneuver through?

Hunter: Definitely, privacy is something that we take very seriously; it’s very important. What it really comes down to is trust. If you take a retailer who has a fair amount of data on an individual household, when you sign up for a loyalty card, there’s a handful of questions that you have to answer, not just your personal identifying information, but how do you want this information used? How do you want the retailer using this information? Most retailers will have that, a question along the lines of, do you mind if we use this for marketing information? Some people say, no they don’t want it to be used and some say yes they do want it to be used. Those that don’t want to be contacted, we don’t use their information so they’re not going to get those special offers. But for those that have elected it to leverage for marketing, they do get value from it. The retailer very much protects that information, the personal identifying information is not seen by very many people. Most of what we work with is just a unique number so to us it’s a discrete data point that allows us to look at individual households, but when it comes time to actually communicate to the person, whether it’s email or mail sometimes maybe text messages, depending on the retailer, then that unique number is attached to the PII information, but only for those people who have elected to be contacted.

Long: John Bailer we go to you for the next question.

Bailer: Thank you. Just a quick follow up. With that data that you’re collecting, not you are collecting, but is being collected, I imagine that there is also combination with other data sources. So you might have identification of some characteristic of a person shopping there. Do you ever try to link that with different data sets maybe from the census or some other sources to try to get other insight?

Hunter: Depending on the business need, the answer to that is yes. It may not be necessarily linked at the individual household level, it might be linked at the individual city or region of the country, but you can definitely begin to get, by what we would call data fusion which is what you’re saying is can we begin to fuse databases together, triangulate databases together, we do indeed look at U.S. government census information that tells us about demographics. There’s other advertising databases that you can begin to look at such as TV or radio. You can also begin to fuse, there’s more and more information coming out on the Internet where you can begin to fuse as well. So data fusion, depending on the business objective, can be quite powerful.

Long: Richard.

Campbell: Paul, this is a little bit different track here, but one of the things that I do when I teach our media survey course, we look at media economics, we look at new products and how hard it is for, in my case, new media products to make it onto the market. I know it’s a very high failure rate for new products in business. Is the kind of thing you’re doing with data helping with that? And talk a little bit about it.

Hunter: New products, we could spend several hours on that, I don’t think time permits that. The most important measure in marketing in my opinion is something called trial and repeat and trial is what percent of people tried your product. I think you can get pretty much anybody to try something once. Trial is the responsibility of marketing and sales, get the product on the shelf, get the word out through mass media or direct marketing so trial is a critical measure that we maintain in the database in terms of how many people tried the product. And then repeat is probably the most important measure within the trial and repeat is how many people tried and then came back and bought it. Depending on the purchase cycle, you need to buy laundry detergent every 64 days, you need to buy toothpaste every 92 days, you need to buy a gallon of milk every two weeks. How often are you coming back buying this particular unique new item? And depending on the category you’re in, we tend to see some repeat measures as lackluster. So it used to be state of the art in the industry with panel data in the 1970s and even today some companies still use it. Trial and repeat measures take maybe seven or eight months to get. With the loyalty card databases and depending on if you look at a set of households that are always in your stores, you can begin to look at trial and repeat in a much more rapid fashion, you don’t have to wait eight months after the product launches. You can look two weeks after the product launches, we can pretty much estimate if the product is going to make it or not. The benefits to that that are many-fold, from a retailer’s perspective they have a finite amount of space. So depending on the purchase cycle of that category, if repeat rates aren’t up there, then the retailer may pull the product and bring something else in. Likewise, the manufacturers can be looking at this data and say, trial is really good, but repeat is not, can we begin to find out why, and maybe do something on our end in improving the product to try and salvage it. We definitely have a huge amount of failure still in new products today, but through statistics, through the metrics that we hold, we are no longer waiting eight months, the window of success is getting down to months, to a handful of months, not a half a year or full year.

Campbell: You said earlier in the program that companies actually learn a lot from failed products and that’s not necessarily a bad thing.

Hunter: That’s true. All the manufacturers out there, the big ones, General Mills, Nestle; they’re committed to new products. They’re looking for the next big billion dollar idea, the next big successful story. And I think they’ve kind of recognized that when they launch things, they may not make it, but they have put in place personnel, R&D department, systems to learn from it. I think it keeps excitement in their categories as well; they have the dedicated sales force. That dedicated sales force needs something exciting to talk about and so hence I think there always will be new products, but the precision and the measurements around whether they are going to make it or not have just really been compressed through time. At the end of the day, I think the manufacturers have to be smarter and sharper when they approach the retailer about what really is a point of differentiation, so how? The other thing we can tell about new products through the data is, and this gets a little bit more sophisticated in the statistics, but we call it source of volume. A retailer is also getting sophisticated with the data and measurements. If you’re bringing me in a new box of cereal the first question a retailer may ask is what’s your point of differentiation and if I put you on my shelf, where’s the volume going to come from? It is going to come from my own label brand, is it going to come from another major brand? What price point are we going to be charging for it? And through the metrics that the loyalty card data provide, even not only can we look at trial and repeat, but we can actually say where it’s coming from. So if the product is going to make it and it has really good repeat rates, we can also through the data say that because it’s going to make it, we can also tell you who you’re going to have to delist because it will cannibalize from the other item.

Long: You’re listening to Stats and Stories and again our topic today, personalizing the world, how businesses are moving from mass communication to highly personalized marketing strategies. I’m Bob Long; our regular panelists are Miami University Statistics Department Chair John Bailer and Media, Journalism and Film Chair Richard Campbell, and our special guest, Paul Hunter, the managing director of customer knowledge of dunnhumby which works with some of the world’s largest retailers and brands. We go back to John Bailer for our next question.

Bailer: Yeah, thanks Bob. Just a quick follow-up there Paul, on this issue of these new products; I mean with new products you’re targeting particular segments; you’d like to impact as large a population as possible. You had said earlier the idea about painting a profile of lifestyle with some of this information you’re using, can you talk a little bit about this idea of market segmentation and how the data you collect can be used to think about sort of the smaller populations that you would customize a product for?

Hunter: Absolutely, segmentation, which is a statistical technique, is something that we thrive on and embrace wholeheartedly within our business. Within marketing, the two extremes are mass and one-to-one and mass communication goes out to everybody and then the one-to-one communication, everybody gets a different message, but segmentation allows us to play to the middle and segmentation is what we really use and we use it for variety of techniques. And depending on the retailer, we may hold up to 80 different segments on one individual. Typically though, the most common is about five to seven segments in the retailer but we use a variety of techniques to get to the segmentation but the most common that we use is cluster analysis. And what clustering does is it says, what are common traits that would allow us to group one set of households versus another? So if you go back and you look at maybe one of the techniques or one of the business initiatives is we’re looking to send a marketing message to a set of households that might be more elderly. Well we hold a set of underlying data points that would say there’s certain items in the store that would indicate that they’re more elderly. So maybe it’s they’re buying certain types of pharmaceutical, not pharmaceutical but over the counter types of drugs. They’re buying certain types of say adult incontinence types of things. That would allow us to okay, here is a group of households that we can begin to cluster on and nobody else is buying these so they’re not the elderly set of households so we can begin to say that through the database we found a very small percentage but maybe it’s three or four percent that we can call the elderly cluster and so then if you’re a retailer that’s trying to win business or reward loyalty to the elderly population through your marketing message, that statistical technique, clustering analysis, allows us to satisfy that need in a much more precise way that’s much more highly personalized than doing mass communication. Likewise, that clustering technique then affords the retailer the opportunity to do one-to-one communication. So you might have two different sets of elderly households; one who’s very loyal and one who’s maybe not as loyal. And so you can begin to intertwine multiple segments together to say that not only do we want to segment elderly households, we want to segment elderly households that are not as loyal and do maybe a type of acquisition marketing initiative. That’s essentially how they’re brought together and we’re a big fan of clustering, we’re a big fan of segmentation.

Long: Paul, I’m kind of wondering, right now you can look at my loyalty card and even age group I’m likely to be in, those kinds of things, but how about the future? Are you able to say down the road, Bob’s going to need, or Richard or John are going to need this. Are we at that point where we can predict shopping trends?

Hunter: Prediction requires forecasting and underlying that is statistics of course. We would like to say that forecasting is an art, not a science. So any type of forecasting will have degrees of error. While we can do it, we’re not as confident or robust in that. In lieu of that, what we have done and can do for retailer or for a client is what we would call “profile their journey through life” and you may not do it for an individual year, you may do it for a group of let’s say, five years. What does the typical family between the ages of 20 and 25, what are some of the categories that they play in? Or depending upon more than let’s say a family between the ages of 25 and 30 and 50 and 55. So we definitely can profile that journey through life and begin to then say, this individual household looks like they’re going to move from one segment, 50-55 to 55-60, these are the types of things that we should be marketing to them differently. So it’s not as precise but we definitely in the clients we work with, they are smart to not just serve them today, but serving their future needs. So that’s one technique that we use.

Long: We’ve got a couple more questions yet before we wrap up with Paul Hunter. Richard Campbell, I’ll go to you next and then back to John Bailer.

Campbell: Paul, one of my obligations in the Media, Journalism and Film department is we train a lot of young journalists and one of the things that’s a challenge for us is having them cover stories that have data and statistics in them because this isn’t their specialty area, but they have to write those kinds of stories. So do you have any kind of advice for journalists or particular pet peeves that you see that aren’t covered as well as they should be in your area.

Hunter: My counsel to your students would be to embrace statistics and embrace data because there’s only going to be more and more of it coming out. Is your question more about in our industry of data mining, things that are my pet peeves? Is that your question?

Campbell: A lot of times the kind of work that everybody does and you do in terms of marketing are only known to the public through journalism because that’s who comes and reports this. So I want to know how well the journalists represent your area and your field and your work and when you’re reading stories that have data in them, are there things that sort of jump out at you as they didn’t get this right?

Hunter: Got it. I would say that the coverage of data mining and managing it for business purposes is not getting as much coverage as I would like to see. That being said, within the coverage that I do see, here is what I personally would like for journalism to uncover more and maybe this is only my pet peeve, using your words, is quality control. We’re all humans; we make mistakes. When you fuse databases or when you look at an analysis typically when it’s an analyst sitting by themselves and they’re under pressure where they have time sensitive material that they have to get out the door, if you had another person looking over their shoulder, would they have caught an error in the code, would you have caught an error in their analysis? I think quality control is an error that could be more uncovered and the techniques and how you get to better quality. That’s me personally and the reason for that is that I’ve been a victim of it. I’ve done it myself and I’ve received poor quality information because things were rushed and people weren’t spending time on it. I think the other area is, going back to my earlier comment that there’s more and more data coming out, is not to be fearful of statistics. At the end of the day, it’s common sense but to embrace statistics and maybe talk a little bit more about the methods, a little bit more about the math and I think our country will only be stronger if we embrace statistics, if we embrace data analysis and mathematics more so than we have in the past because it’s where the future’s going to be and it’s not just data mining, it’s all industries that have reams and reams and reams of data that are going to need to be analyzed. So for journalism to actually be brave and explore and explain to the common person a methodology such as regression analysis or clustering analysis. That would be kind of cool.

Campbell: Thank you.

Long: John Bailer, we have time for a quick question from you.

Bailer: Sure, absolutely, I was going to challenge you on the idea of embracing statistics but I think I’m going to let that pass. The choir echoes ahem. My last question for you, what kind of skills, quantitative and otherwise, would someone need to be able to contribute in your field?

Hunter: The first one is I think a skill that some of us have and some don’t but curiosity, it’s a natural desire to ask questions about why or how and because if you’re not curious you can quickly get bored with the data and to be naturally curious is the first one. I think obviously a nice mix of working with computers, both personal computers and midrange computers, Linux or Unix, not to be intimidated by the computer platforms and then recognizing that there’s certain statistical packages that you may have to use, whether that’s R or something else of that nature that allows you to mine very large databases. So that is probably three quick skillsets, is curiosity, computer skills and analytical skills, but probably the most important is what I would call to be able to walk and talk and chew gum at the same time and that’s to take the insights and tell a story. Why does it matter? What do I do differently as a business? And that’s the harder one to find.

Long: Paul Hunter of dunnhumny, thank you so much for joining us on Stats and Stories today.

Hunter: Thank you.

Long: And if you’d like to share your thoughts on our program send your emails to StatsandStories@MiamiOH.edu. Be sure to listen for future editions of Stats and Stories where we’ll discuss the statistics behind the stories and the stories behind the statistics.