[MUSIC] So now just to recap, we've done two studies that I think shows some very, very interesting findings about spreading contagion influential users. We've looked at demand for netgrocer.com spreading through neighborhoods throughout the United States. We've also worked out who's influential for whom in social networking sites. So we've been in the real world and in the virtual world, now let's flip back to the real world again. So one area that gets looked at a lot for influencing contagion is among professionals, particularly professionals who are in the medical community. So what I'm showing you here is a diagram. It's coming from a paper that was written by two of my colleagues here at the Wharton School, and also one of their colleagues out at the University of Southern California. This is a network, you might recognize this kind of diagram from the very first thing that we looked at when we started this discussion about networks and neighborhoods. That YouTube link that I gave you to show the spreading of obesity, that controversial study, throughout Boston and the United States. This is the same thing here. What this is showing is among 174 physicians in the Los Angeles area, who is connected to whom, and the size of the circle is also indicating some degree of influence. So what the firm, the medical firm, and the researchers wanted to understand is who in this network is influential for whom and by how much. So again, to go back to my all-purpose friend Chris who's now going to wear the hat of the doctor. If Chris is a Physician in L.A. and I'm also a physician, if he starts prescribing a certain drug to his patients, maybe I'm going to follow along and do the same thing. Now of course if the drug company knows which doctors are influential, that's very important for targeting purposes. So what the colleagues did is they used the methodology very, very similar to what I showed you earlier, remember we had the four zip codes? Z1, Z2, Z3, Z4, in this case its the same idea, but instead of zip codes they're doctors. So you looked at which doctors were connected to whom, that's the first piece of information. And then secondly, which doctors already prescribe the drug and which doctors had yet to do so. So remember when we study contagion processes we need to know two things, who's connected to whom and who's done what up until the current point in time. Now, they found a couple or really interesting things, let me give you the highlights. They measured influence and contagion in two very different ways. One way was to just ask people on a self reported basis, hey, you're influential on a scale of one to ten. Well it turns out if people say they're influential it's not too bad but it's actually a relatively weak predictor, compared to an indirect measure of measuring influence, which is whether or not I'm citing Chris' work as a doctor, I'm referring to a scientific study. So instead of looking at a measure of influence that is self-reported, what they did is they had another measure of influence that was in some sense more objective. Were doctors referring to each other's work and each other's scientific studies? And they found the second one was more important in predicting the way these drugs were going to diffuse. So now let me give you the main takeaways from the study which I think was really, really fascinating. So first of all, the firm found that was really helpful for them to try and understand the network structure of the customers in this case, the doctors. And also an understanding in this network structure, they were able to identify that a contagion process was at work and the contagion process was driven by these influential people. Now what's really interesting about this is some of the influential people weren't necessarily the people who put their hands up and said, hey, I'm influential. But they were the people that they figured out indirectly were influential. That is the doctors to whom others referred in terms of scientific studies, and citations, and so on. And they are also some quite special people who had their feet in different camps, as it were. So, in the study, they found that there were certain Asian-American doctors who both were influential for other Asian-Americans, but also for people outside of their ethnic group, as well. So, very, very interesting. It tells us that understanding the "network structure" is important. Number two the contagion occurs through the network, and number three in all networks there are certain special people who are more influential than others and our job as marketers is to try and understand who those are. So now let's turn to our fourth and final study. So we just finished a study of physicians in the real world, we've looked at some other things in the virtual world. And now we're going to go back to the virtual world again, but with a real world twist. The company that we looked at in this case is a company called Bonobos. It's been around since about 2007, selling mens clothing online. Also selling though traditional retailers and I think I mentioned them a little bit earlier in the piece as well. This resulted in a paper that my colleague Jae Young Lee and I wrote about something called neighborhood social capital and online sales. So let's look and see what that is talking about. So now I'm just showing you a screenshot of the company Bonobos. So you can see that their selling to men, the target is males aged roughly 20 to 45, who is somewhat fashion forward and looking for affordable, fashionable clothing. So, what we wanted to do in this case is, we wanted to try and understand whether or not there was real world interaction that was increasing the virtual world sales of this company. So, what I mean by that is, that my friend Chris and I, so Chris is back in the picture, this time he's just a regular friend who wants a pair of pants, he's not a doctor any more. So Chris and I are friends, and if Chris happens to buy some items of clothing from bonobos.com and I see him wearing them, and he tells me about them, is that going to lead me to then increase the chance that I buy from the same website. That's what we wanted to look at, whether or not interaction in the real world was going to lead to additional sales in the virtual world or the website of the company. That was the first piece. Now the second piece, again here it's on the slide, we wanted to see whether or not there was an effect of something called social capital. Social capital is a really fascinating concept and it's one that was really coined, I believe, by a fellow at the Harvard University's Kennedy School of Government, a gentleman by the name of Robert Putnam. Who wrote a book called Bowling Alone. You know, think of bowling temp and bowling in America, the metaphor is bowling alone that people were maybe somewhat less social then they used to be. Maybe we're spending all of our time online, we're disconnected from other people. And so what he wanted to do is he wanted to understand in local neighborhoods how connected people were to others. Did they participate in churches and tennis clubs and get together with each other and so on? And what he did is he did a huge survey called the Social Capital Community Benchmark Survey, where he and his team literally went around to about 30,000 different zip codes within the United States. And they ask people like my friend Chris, hey Chris, do you like your neighbors, do you trust your neighbors, do you interact with your neighbors? And so, some very interesting data was collected about trust and interaction. So neighborhoods with more trust and interaction, and neighborhoods that had higher social capital, we wanted to see whether neighborhoods with more social capital, there would be more sharing, more efficient sharing of information. Now, I just want to draw your attention to one other thing about this particular website. There are three conditions related to the product that are particularly important for our study. The first is that items of clothing, like the sweater that I'm wearing, have what are called non-digital attributes. What does that mean? Well a non-digital attribute is something that's very hard to represent perfectly over the internet. So price is a digital attribute. If I go to Amazon and I see a book is costing $20 I know it's $20. It's easy to communicate price information over the Internet, just as it would be if it were positioned in a store. However to try and communicate how this fits in fields is actually quite difficult. So in that case transmission of information from one customer from one customer to another, in the real offline environment could be very, very important. Secondly, in our study, we wanted to focus on those customers who hadn't yet bought anything from the website. This was going to be their first time purchase. Why is that? Well, because once you bought something from a website and you've tried on the sweeter, you have your own judgement. You don't necessarily need the opinion of other people, unless it's about the overall fashionability. And then finally this is a product that is socially visible, so it might be one that actually generates a conversation. So I might see Chris and say, hey Chris you look very well dressed today, where did you get those pants? And then a conversation ensues. So I just want to reiterate that this is a product category that's a little bit different to most of the products that we've been talking about that are being sold by our friends at [INAUDIBLE] Places like soap.com and diapers.com. Those products have primarily digital attributes. There's not a real surprise if you order some Tide detergent and it shows up at your house, you know exactly what you're going to get. There's no problem of communicating that through the internet. So this time we wanted to look at the business that was little bit different. That was a fashion business that had this other properties. So let me know what the raw of data looked like. These are just the sales data for the company over the first, I think 42 months or three and a half years of operations. You can see that over time the number of new customers coming in is going up. You can also see with the blue arrows that in neighbourhoods where there is more trust and interaction, the sales are higher than the neighbourhoods where there is less trust and interaction. So, what does this all mean? Jay and I put together a statistical model to try and understand this in more detail and what did we find? And the findings are here shown on the screen, but let me explain what's going on here. We found that of about the 6,000 trials that we looked at, at least half of them were influenced by what we call social learning, meaning the evidence from the statistical analysis suggested that some of these new customers became customers in their local neighborhood told them about it. So that's a pretty important effect, half of all sales of this company. Secondly and related to that, we found that the customers who came later on, were the ones that were most influenced by the social interaction. This ties back to some of the themes that we talked about earlier. The people who do things right in the beginning, they don't usually need to rely so much on the opinions of others, they just like to go out and do stuff. The people who come in later, they require more social information typically. And that was also confirmed in our study. The second thing that we found, that was really I think the most interesting finding to us was the following, in neighborhoods where there's more trust and interaction, more social capital, there's not necessarily higher sales. So it's not that just neighborhoods with trust and interaction have people who buy more stuff. But what happens is, in those neighborhoods, when information gets transmitted, it's more believable and it's more trustworthy and more efficient. So if Chris and I live in a neighborhood where we trust each other and like each other. If he tells me something, I put more weight on it. That's the result that was coming through here. So, how could the firm Bonobos.com or any firm kind of use this information? Well, when we did our analysis we were a little bit restricted to only the zip codes where the social capital survey measures are being collected. So there are many, many zip codes in the United States for which those measures were not collected. Now, earlier I think I said that Mr. Putnam went out and he measured 30,000 zip codes, actually just to be clear, he measured 30,000 people, who were living in about 1,000 zip codes. So if a firm really wanted to use this, clearly knowing only about 1,000 zip codes is not quite enough. So here's a question I want to put to you all out there, and then I'll give you the answer. If you could think of a proxy, that means some other variable other than the true measure of social capital, that would indicate that males age roughly 20 to 45 were socialized together and had some level of social capital, what might it be? Number of hospitals per zip code. Number of churches, maybe. Number of rugby clubs, okay? You're going in the right direction. Turned out that the number of bars and liquor stores per capita was a very nice predictor of the efficient diffusion of information among this group. Why am I telling you this? Because I want you to be creative and to think a little bit expansively when you start to use these concepts, and you to start to think about, gee, how could I use this idea for my own business that I'm working on, or the company that I'm working at now? So that brings the conclusion to this piece of our discussion. I hope you enjoyed those four studies. Number one, the netgrosser.com, number two, the social networking site of influence, number three, looking at the fusion of drug prescribing behavior by physicians, and finally, how offline interaction is affecting people's sales of product on the Internet for bonobos.com. [MUSIC]
