4. Integrating Online and Offline Retailing: An Interview with Peter Fader and Wendy Moe
Although I have spent most of my career studying the click-click-click of shopping carts in physical retail stores, the click-click-click of online retailing has emerged as an important new window on shopper behavior. With new technologies moving into retail stores, and with the increasing integration of online and offline commerce, studying online retailing also provides insight into how shopping might evolve in the future. I had long been aware of Wharton School professor Peter Fader’s shopper modeling studies in physical retail stores, and I had a chance to collaborate with him in studying the in-store shopping process. In the late 1980s and early 1990s, he and other researchers used a growing avalanche of point-of-sale scanner data to analyze what people buy. Their models could help us understand, for example, why shoppers bought one brand of orange juice over another. Although linking sales data to specific customers was a leap forward, it still offered limited insight into the behavior of shoppers in the store. Our work initially helped fill that gap in the physical retail space, by carefully observing and analyzing how shoppers shop.
But then came the Internet. In addition to his studies and his modeling of physical retailing, Fader, along with his colleague Wendy Moe, an associate professor of marketing at the University of Maryland, conducted path-breaking studies of online behavior. In studying their work on online shopping, I saw that the core principles bore a striking resemblance to what we see in physical stores. The point-by-point clickstreams of online shoppers are similar to the point-by-point visiting, shopping, and purchasing behavior of in-store shoppers. Although I focused more on “crowd” statistics in my work, looking for descriptive insights for all shoppers or major groups, Fader and Moe (along with other colleagues) looked at patterns of individual shopper behavior as a means of assessing the drivers of individual behavior. The picture is richer from both perspectives than from either alone, and each confirms the other. In this interview, they share some of the insights from their research.
How Did the Internet Change the Study of Shopping Behavior?
Fader: For my first ten years or so here at Wharton, I was looking strictly at point-of-sale scanner data, just analyzing what people buy and completely ignoring the context around it. Not that I wasn’t interested in the context, but there was just no data. So, we had all these models trying to help us understand why people buy this brand of orange juice instead of that. The models worked great and people applied them to very different areas such as pharmaceuticals and financial services. Then everything changed with the dot.com revolution. As everyone was jumping into these uncharted waters, I initially wanted no part of it. I figured the process of someone standing at a shelf and deciding what juice to buy is going to be very different than someone sitting at the computer clicking through a bunch of different books or CDs—until I actually looked at the data, and it turned out that the patterns were remarkably similar. The same types of patterns were apparent in both purchasing processes. It was really fascinating. I never expected it.
Moe: I focused on online research from the beginning. My early research looked at online shopping behavior, how frequently people come to the store, how much they visit it, and what kind of search activities they use in comparison to other stores. That can give us an idea of how much they purchase. Then we can forecast their purchasing. We looked at patterns across visits and repeat visits to predict what they might buy in the future. I realized, however, that for some categories of products, consumers are not necessarily going to make repeat visits before they buy, so I focused on page-to-page behaviors within a single store visit. I looked at issues involving the focus of search behavior: Does one look at a specific brand or jump around across categories? Within a category? Is the consumer buying or just browsing? What do these patterns mean for purchase behavior?
E-commerce marketers have an abundance of data that most offline retailers don’t have access to. They use this for better diagnostics and more accurate forecasting. They also can experiment with product layout and promotional messages. In the physical world, customer data have been focused mainly on when, what, and how much people buy. What researchers in physical stores haven’t been able to measure as well are activities such as comparison-shopping and information gathering, which often have a strong influence on the final choice. Insights from online data might help offline retailers better understand shopping behavior.
“I figured the process of someone standing at a shelf and deciding what juice to buy is going to be very different than someone sitting at the computer clicking through a bunch of different books or CDs—until I actually looked at the data, and it turned out that the patterns were remarkably similar.” —Peter Fader
In What Way Are the Online and Offline Patterns Similar?
Fader: We found that people’s tendencies to do something and decide whether or when to do it again—and how many more times to do it—were similar. The mathematical models describing the behavior could be applied in almost a “cookie cutter” way to a variety of products, from cans of baked beans on the shelf at the Safeway to books online at Amazon.com. I threw myself into e-commerce because there we could see not only what people buy, but also the process leading up to it. We began looking at the interplay between visits and purchases online. There were all these patterns, that we could never see before, that were consistent with what we had been seeing. Then we came across your PathTracker studies in bricks-and-mortar stores, using RFID tags on shopping carts, and this allowed us to integrate these two areas of research, measuring shopper behavior with a variety of different tools (see the sidebar, “Studying the Same Shoppers on Different Paths”). We could take the rich context that we were able to see in the online world and marry it with our deep understanding of what happens in the grocery world. It allowed us to address questions such as: Which zone will you visit next? Are you just passing through or are you really shopping there? And if you’re shopping there, what things, if any, are you going to buy?
Studying the Same Shoppers on Different Paths
Retail shopping: Radio frequency identification (RFID) tags underneath shopping carts track consumers’ in-store movements.
Eye tracking: Researchers ask subjects to view print advertisements and capture the subjects’ eye movements using infrared corneal reflection technology, or use the same technology in the store.
Web browsing clickstreams: Consumers’ web-browsing patterns are tracked by recording the sequence of web pages visited in a session.
Information acceleration: Researchers immerse subjects in a multimedia environment to understand how they collect information and make decisions about a radically new product or service.1
How Are Paths in the Supermarket Similar to Paths Online?
Fader: Online and offline marketing both have path-type data associated with them, yet very few people consider the path aspects. They simply look at outcomes. Look at some of the e-commerce work on how people click from page to page and whether they are buying. I tended to think, like most people, that there’s nothing else like it in the physical world. The Internet is brand new, right? But clicking through an online site is a lot like someone pushing a cart up and down the aisles. Turning left or right in the grocery store is a lot like clicking on this link or that link, even though the physical movements are quite different. The decision process, when you get down deep, appears to be pretty similar. A path is a path, whether it’s online or offline, people or birds.
Can Online Retailers Learn from Offline Shopper Behavior?
Fader: Take a small example: Herb, your in-store research has found that shoppers tend to look to the left and move to the left, counterclockwise, as they move through the store. Where is the shopping cart located on just about every online site? On the right. You’ve got people who will experiment with absolutely everything in designing their website—different content, different colors, different everything. But the shopping cart is always on the right. They have no idea why. Should they test the cart on the left?
We should no longer be surprised when we see these online/offline analogies. We should expect them. But a lot of people don’t buy it. They say, oh no, come on, looking at a screen is totally different than pushing a cart in a store. There’s very little evidence to suggest that it’s really different. On the other hand, offline retailers, who have spent all this time laying out the aisles and getting the merchandise at the right height, don’t think they could learn something from online. But I believe they can. The ultimate irony is that bricks-and-mortar retailers are outsourcing their online business to online firms. They say: Make our website as efficient as possible and just tell us how much money we are making. These online companies are running thousands of different experiments every day. But they tend to be two completely different species of people, and they are not taking any of these learnings back to the offline store.
Some of the areas that we have studied that have implications for both online and offline retailing are crowding and herding, sequencing or licensing (buying “vice” products such as chocolate after “virtue” products such as vegetables, for example), shoppers speeding up as they move toward their goals, shopping momentum, the impact of variety, and hedonic shopping behavior.2
Tell Me about What You’ve Found Out about Crowd Behavior?
Fader: Studies online and offline show similarities in crowding and herding. It boils down to this: If I’m in a store and I see a big crowd of people down an aisle, does it attract me or repel me? There are two schools of thought. One says that a crowd attracts people, the other that it repels. But it’s not that straightforward. People tend to go where there is a crowd—but they won’t necessarily shop there. I see a crowd in a store, so I am going to push my cart down there to see what is going on; however, it might just be too crowded to engage with products on the shelf, so I’m going to move on. What we found is that it doesn’t just depend on the individual but also on the type of behavior. We have some insights, but there is still a question about how these two forces counterbalance each other. This is an excellent example of why we need a statistical model.
It is also worth mentioning the connection here with GRPs (gross rating points) and their constituent components, reach and frequency. The question is: Do we want to get people staring at the shelf often, or do we want to get a lot of people to stare at the shelf? You might get a lot of reach but repel people at the same time because of the crowding.
What Have You Learned about Licensing and Sequencing—Such as the Purchase of Vice Items After Virtue Items?
Fader: There is a sequence in how people buy things. One important driver of this sequencing may be “licensing” behavior. If shoppers buy a virtue product—something good for them—then it gives them license to buy a vice product—something that they might enjoy but is probably not good for them. If they pick up broccoli or tofu in the produce section, they can buy the ice cream or chocolate cake or cigarettes. It has tremendous implications for the placement of products on the path through the retail store or online. This is one of the reasons why sequence matters. The chocolate cake at the beginning of the shopping trip might be viewed differently than at the end, when the shopper has accumulated some virtue products and now might be more willing to indulge in a vice.
We see this in the laboratory, but it has yet to be validated in the real world. We sit people down and show them a list of things and ask them to indicate which ones they would buy. But for certain people, you prime them by saying you’ve already bought a certain product (virtue or vice), and then see if they buy the vice product after the virtue. There is some variation across people in how attractive the vice products become after the virtue ones.
“If shoppers buy a virtue product—something good for them—then it gives them license to buy a vice product.” —Peter Fader
What Have You Found Out about the Pace of the Shopping Trip?
Fader: Research has shown that the closer you get to your goal, the more you speed toward it. This is called the “goal-gradient hypothesis.” We see this in the context of loyalty programs, where people buy ten and get one free. A number of researchers have noticed that as you get closer to the goal, you speed up purchasing. This is consistent with what you’ve seen in the store, the concept of the “checkout magnet,” where shoppers move more quickly as they get closer to the checkout. We could extend the goal gradient hypothesis beyond the checkout, to look at other goals that occur as you shop. They might be items on your shopping list, parts of the store that you go to on every visit—for example, to see if meat is on sale each week. This is difficult to study because, other than the checkout, shoppers have different goals. But we believe we will see evidence that this goal-gradient hypothesis applies to intermediate goals as well as the final checkout.
What Have You Learned about Shopping Momentum?
Fader: The idea here is that as more purchases are made, everything in the store becomes more attractive. Once shoppers pick up a number of items, it gives them the momentum to buy more. Once you have two or three items in your cart, you start really rolling and then pick up a lot of stuff. The more you buy, the more you buy. As your studies have shown, the most common number of items purchased in a grocery store is one. That means many shoppers never really get rolling. They may get that one item and, before they can be enticed to buying a second or third, they leave. If they do build momentum, they buy a lot more.
Just as with the other forces in play in the store, the impact of the forces at play such as shopping momentum will vary across individuals. That, again, is why we need a proper statistical model to measure the effect of these forces on individuals.
What Have You Learned about the Role of Variety in Shopping?
Fader: There are some people who say that people like variety. If there’s greater variety, then you’re meeting people’s needs in a better way and, therefore, the category as a whole becomes more attractive. On the other hand, if you have too many forms and flavors, as Barry Schwartz points out in The Paradox of Choice,3 people are actually put off by too much choice.
We’ve only tested this so far in an indirect way—across categories, which is not the best way. We looked at categories that have high variety and those with low variety to see how attractiveness varies across them. What we see is that high variety is more attractive.
This doesn’t necessarily refute your perspective or Barry Schwartz’s that offering more limited selections can increase sales. We’re looking across categories within stores, rather than the same category across stores, and there is tremendous opportunity for reverse causality here. It could be that if a category is more attractive, people really like it, and then retailers are going to want to stock more SKUs in it. That could explain why greater variety in a category is associated with greater attractiveness. There is a lot of testing that needs to be done.
What Have You Learned about Efficiency? Is It Better to Allow Shoppers to Get Quickly In and Out of the Store, or Should Retailers Try to Prolong the Trip?
Fader: Traveling salespeople are famously well organized. They have to be: They are always on the move, visiting a certain number of clients every day, so they need to find the most efficient route possible. Our studies of the paths taken by 1,000 grocery shoppers at a store in the western U.S. have found that shoppers, unlike traveling salespeople, are often quite inefficient.4 They might choose the right order in which to travel to find the products they want, but they take too long and go further off course than they need to. Another interesting finding is that inefficient shoppers tend to have more in their carts than those who shop more efficiently, so this inefficiency within reason is not necessarily a problem for retailers.
This research has major implications for both store owners and brand manufacturers. Retailers want their customers to have as efficient an experience as possible. On the other hand, they want the shopper to stay longer and interact with more products in the hope that it will drive more impulse purchases and incremental revenue and build the relationship that will make shoppers want to return. If, however, they prolong the shopping trip through confusing store design and bad signage, customers will get annoyed and not return. More time spent shopping could be a good thing if it is a sign of increased engagement but might be negative if it reflects confusion and aggravation.
In the online world, measurement has taken a radical turn from looking at how many unique visitors are attracted to a website to measuring time spent per session. Again, there is the divergence between efficiency and engagement: To what extent should retailers, whether online or offline, help shoppers finish shopping as quickly as possible or try to hold them for as long as possible?
Moe: I conducted a study on the impact of pop-up promotions on people’s online surfing behavior. People often complain that these advertisements are annoying. They don’t like them. But actually they influence their behavior positively from the perspective of the retailer. People who are served pop-ups at the right moment actually stay on the website longer and shop and search a little bit more. If the pop-up itself has good content that matches their needs, visitors are encouraged to stay, search, and buy. If the pop-up is on a gateway page—a home page or category page—that visitors use to get to the products they want to see, a pop-up is stopping them from getting to their ultimate destination. These are received poorly.
“People often complain that these [pop-up]advertisements are annoying. They don’t like them. But actually they influence their behavior positively from the perspective of the retailer.” —Wendy Moe
This Raises the Question of Whether Shoppers Are in the Store for Utilitarian Reasons Alone or If They Are Interested in an Experience. What Is the Difference?
Fader: Shopping can be for a utilitarian purpose—something that has to be done—or it can be done for a hedonic purpose—for the sheer enjoyment of it. Online grocery shopping has not caught on in the U.S. to the same extent as the U.K. This might be because larger U.S. bricks-and-mortar stores offer the hedonic experience that online shopping lacks. Many Americans live in large houses spaced farther apart than their European counterparts, which makes going to the store more of a social experience. Again, this is an area ripe for investigation in both the online and offline world. In the online world, you can watch the same individuals over a number of shopping trips and start to notice patterns. Offline, what is needed is to marry data from a series of PathTracker5 studies over time with data from a shopper loyalty card to find out exactly who is doing the shopping. This would show how often they are shopping hedonically versus in a utilitarian fashion and whether there were patterns involved.
What Have You Learned so far about What Shoppers Are Looking for When They Go Online?
Moe: My research has looked at the underlying objectives of online shoppers and the expression of these objectives in purchases. I identified four distinct types of visits associated with different online behaviors, as follows:
Directed-purchase visits are where the shopper enters the store with a clearly defined purpose of walking out with a specific purchase.
Search/deliberation visits are utilitarian visits where a future purchase is being considered, and the store visit is designed to gather relevant information before buying.
Hedonic browsing visits are more about enjoyment, where shoppers browse without looking for anything specific but might make an impulse buy.
Knowledge-building visits are also enjoyable but are more geared toward collecting information for possible future purchases.6
It is important for online retailers to understand this to target their marketing activities effectively to the right people. Shoppers come into bricks-and-mortar stores for some of the same reasons.
How Do Online Retailers Use These Insights about Shopper Visits?
Moe: The next stage of research looks at differentiating between online shoppers not just according to what pages they are looking at, but by also actually examining the products they are interested in.7 In other words, what are the characteristics of the products they are searching for and interested in? And what are their ideal products? Building a model based on data from this research enables the retailer to estimate the probability of purchasing. For example, if someone looks only at a series of black shoes, you can infer that she has a clear preference for this color shoe. Someone else might be looking only at shoes in a certain price range. The sequence of pages tells the researcher something about what a person’s preferences are. This helps predict not only whether they will buy, but also what.
By understanding better what shoppers are looking for, retailers can, in theory, create a virtual smart salesperson to help. This assistant might be compared to the salesperson in a physical store who, through observation and experience, can help shoppers find products they prefer while carefully screening out items they feel the shopper is uninterested in buying. A model of purchase expression could help create a virtual assistant that could do the same thing.
This Captures the Whole Point of What We’ve Called “Active Retailing.” Online Is Leading Offline in This Area. How Does This Come into the Physical Store?
Fader: This has obvious implications for online retailing, but as more interactive technology comes into offline retail stores, through cell phones, PDAs, or other devices, it could also be done in the bricks-and-mortar world. To offer this kind of assistance, you need to understand shopper behavior and how it relates to purchases.
How Do Some of the Complex Forces of Shopping Behavior Play Out? Why Is There a Need for Better Modeling?
Fader: As we’ve discussed, there are sometimes countervailing forces in shopping. Crowds attract shoppers but might make them less likely to actually shop. (This is similar to the attraction of the Long Tail discussed earlier, which attracts shoppers to the store because they know they can get anything they need—although they may only buy from the Big Head.) The checkout serves as a magnet to draw shoppers to the end of their journey, the goal gradient, but at the same time, shopping momentum makes shoppers absorbed in the process of shopping the more they shop, spending more time in the store. Efficiency is another area of balance. On the one hand, you want the trip to be as efficient as possible, so the shopper finds what he or she needs and leaves. On the other hand, you want to create engagement, to make the shopper stay longer and interact to drive more impulse purchases and form some kind of relationship. You also want variety, but not too much.
These forces counterbalance each other, which is why we need a statistical model to understand behavior. There is no way we can just look at the observed data and figure this out. These effects vary across people, and their interactions also vary across people. We need a proper statistical model that lets each person have his own momentum effect and each person have his own checkout attraction and to see if we can pull him out from the data.
What Topics Are You Studying Now?
Fader: A big issue we are looking at is edge detection in the “stores within stores” in what you call a “compound store.” Since the edges of these stores are not always formally delineated, we are defining the way shoppers see different parts of the store. Where are the invisible walls? For example, if people tend to circulate within one area, it could be a self-enclosed zone. Why is that? How do they move beyond the borders of that area? Is it tied to products, or is there a psychological reason? We can study this through eye tracking and through models drawn from disease mapping, which look for clusters of diseases. There is a lot of scattered literature about how we can do this kind of boundary definition or edge detection, and we’re just now starting to apply that to the grocery store.
Moe: I’m trying to refine the model for a virtual salesperson, as we discussed, and also looking at the role of online reviews. There is a lot of research that shows reviews have a significant impact on sales. If you have more positive reviews, or even just a higher volume, you get more sales. But the process of posting is something that we don’t understand very well, and lots of managers and marketers and some researchers have speculated that there are biases in those reviews. Posted reviews tend to present more extreme views, so they don’t really reflect the true quality of the product. I’m trying to separate the effect of that bias from the effect of true product quality. The purpose of doing that is that some marketers have started to seed some of these chat rooms and product reviews with their own comments to try and get the ball rolling. So the question is: Do these fake marketing posts have the same effect as an organic consumer-posted review?
Review Questions
1. In what way are the online and offline shopper behavior patterns similar? What are the similarities between online and offline shoppers’ patterns?
2. Describe licensing and sequencing. How are these concepts related to navigation both online and in bricks stores?
3. Consider the issues of variety and range. What have you learned throughout this book about different perspectives on these issues?
4. Discuss the concept of shopper efficiency. To what extent should retailers (online or offline) help shoppers finish shopping as quickly as possible or try to hold on to them for as long as possible? What are potential drawbacks of having too efficient or too inefficient shoppers?
5. Describe how the four types of online shopping visits Moe identified relate to the four shopping states discussed in Chapter 1.
6. What are the next big things in researching online and offline retailing, according to our experts?
Endnotes
1. Hui, Sam K., Peter S. Fader, and Eric T. Bradlow (2009), “Path Data in Marketing: An Integrative Framework and Prospectus for Model-Building,” Marketing Science, 28(2), pp. 320–335.
2. Hui, Sam K., Eric T. Bradlow, and Peter S. Fader (2009), “Testing Behavioral Hypotheses Using an Integrated Model of Grocery Store Shopping Path and Purchase Behavior,” Journal of Consumer Research, Vol. 36(3), p. 478–493.
3. Schwartz, Barry. (2005), The Paradox of Choice: Why More Is Less, New York: Harper Perennial.
4. Hui, Sam K., Peter S. Fader, and Eric T. Bradlow (2009), “The Traveling Salesman Goes Shopping: The Systematic Deviations of Grocery Paths from TSPOptimality,” Marketing Science. See also (2006) “The ‘Traveling Salesman’ Goes Shopping: The Efficiency of Purchasing Patterns in the Grocery Store,” http://knowledge.wharton.upenn.edu/article.cfm?articleid=1608.
5. Moe, Wendy (2006), “A Field Experiment Assessing the Interruption Effect of Pop-Up Promotions,” Journal of Interactive Marketing, 20 (1), 34–44.
6. Moe, Wendy W. (2003), “Buying, Searching, or Browsing: Differentiating between Online Shoppers Using In-Store Navigational Clickstream,” Journal of Consumer Psychology, 13 (1 and 2), 29–40.
7. Moe, Wendy (2006), “An Empirical Two-Stage Choice Model with Decision Rules Applied to Internet Clickstream Data,” Journal of Marketing Research, 43 (4), 680–692.