The Traveling Influence Problem


Penguins move swiftly in water, but slow on land. Influence has similar peculiarities. If you show your new smartphone to colleagues they want to buy it right away, but somehow you mother reacts differently. Or try to start a Mexican wave in a shopping mall: at times it easy to influence others, but sometime it just isn’t. Two articles in the book “The Connected Costumer” try to identify the mechanisms of traveling influence in costumer networks and offer some captivating insights. Let me try to take you along.

It is not a brand-new idea to look at markets as networks and to study the traveling network problem. On the contrary. In my last post I discussed the idea of “opinion leaders”. This idea dates from the 1940ies. In a study about the 1940 presidential elections (won by Roosevelt), the researchers Katz, Lazarsfeld and Merton found that radio messages had little direct effect on the voter’s opinions. Rather, there was a two-step process. Radio messages influenced opinion leaders first and they spread ideas to the rest of the population. So the theory was that influence took a detour. Possibly, marketeers should, in crafting marketing messages, target opinion leaders rather than the general public. As I said in my previous post about the connected costumer, the body of evidence around this ‘opinion leaders’ idea is not definite. One reason is that we do not know how these leaders manage to exert their influence.

In the chapter “The Shadow of Other People”, Ronald Burt tries to unravel two influence mechanisms:  “Socialization” and “Social comparison”.  People spread influence through Socialization when they talk to one and another.  The social comparison mechanism does not need contact. Most people compare themselves to others and are affected by that. An everyday example could be the way fashion spreads. We like to wear what we see (some) others wearing. Often, socialization and social comparison go together: we compare ourselves the most to the people we socialize with. But to solve the traveling influence problem it is important to pull the two mechanisms apart. Burt tries to do so by identifying the network structures that allow socialization and social comparison to take place. He looks at connectivity (to predict socialization) and network equivalence (to predict social comparison).

Let me explain the difference between these two network structures. The figure below shows the network within an imaginary company. The lines show connectivity: two people who are connected by a line, such as production workers 8 and 9, have frequent contact. Two people have an equivalent network when they have an identical pattern of relationships.  Within this company there are no direct connections between the three salespeople (11,12, 13), but they do have high network equivalence. The department managers (2,3,4) have contact with the general manager, but less equivalent networks. Thus, simply put, person 8 and 10 in this company will influence each other by socialization and social comparison, 3 and 1 only by socialization and 11 and 12 only by social comparison.

network of an imanignary company

But is it easier to exert influence through network equivalence or through connectivity? Ronald Burt discovered that network equivalence is the most important cause. It turns out that similar people, influence one another easily – even when they are not connected. The sales agents need no connectivity to influence one another:  they have equivalent networks. But people who have very different networks, such as the general manager (2) and the marketing manager (3), do not influence one another even when they are connected. Connectivity has no influence on them either. It may be counterintuitive, but in the example, salespeople (11 and 12) influence each other more than the sales- and general manager (4 and 1), while the salespeople meet hardly and the managers all the time. However, connectivity counts a lot when people are “near-peers” – thus have intermediate network equivalence, such as the production manager (3) and production workers (8). For near-peers it turns out that the more contact they have the more they influence each other. Thus: network equivalence moderates the relationship between connectivity and influence: socialization helps to spread ideas but only among moderately similar people.

While Burt’s studies give an interesting insight in how influence spreads, they do not address the issue of opinion leadership directly. This is an issue which Jacob Goldenberg, Sangman Han, and Donald R. Lehmann discuss in their article “Social Connectivity, Opinion Leadership and Diffusion”.  Their main concern is how product adoption takes place in a market, and in particular how it bridges the “diffusion gap” between early and late adopters in the market. Often, late adopters have different concerns about new technology. Therefore, once the market for early adopters becomes saturated, the adoption process slows. Late adopters do not naturally follow up on the products early adopters started using.  Can we explain this with network theory? Is it possible to identify the role of opinion leaders in this process?

Since the Katz, Lazerfeld and Merton study, many researchers tried to identify opinion leaders. Here are some traits researchers have found: opinion leaders are more involved with, and have more knowledge of, the products they use, they spend time shopping and have general market expertise, they tend to initiate discussion with consumers and offer market information, they are product experts and innovative, they have more product category knowledge than others, they are confident in product choices and tend to be loyal to brands. So they are super humans! Opinion leaders are totally unlike other people. And, we just learned from Ronald Burt’s study, that people, who are unlike others, are not so likely to spread influence at all. So something is wrong. Goldenberg and his colleagues suggest there may be more, different types of, players involved who have all been called opinion leaders so far. Maybe it is possible to pull those apart.

So Goldenberg and his colleagues look at three groups of traits that help spreading influence: good communication skills, expertise about products and connectivity. One person could have all these traits, but it is more likely that these traits are represented in the network by different (connected) players. Goldenberg and his colleagues try to unravel this by looking at social hubs. In most networks there are many nodes with few connections and a few nodes with a lot of connections (following the Barabási–Albert model). Social hubs have many connections. It turns out that social hubs play an important role in product adoption, and in particular in bridging the diffusion gap. There are two closely related reasons for this role.  First, because of their connectedness, social hubs are attractive information sources. They are connected to many experts, so they have access to a lot of knowledge, but they are not experts themselves and speak the language of novices. This way social hubs can act as information brokers. Second, social hubs tend to adopt early without being early adopters. Early adopters tend try new products first, because of some personality trait. They look out for new products and want to be among the first to try them. Social hubs try new products before the majority because they know about them earlier. So compared to the majority they adopt early but, their product needs are similar to those of the majority.

Is this model the final answer to the traveling influence problem? The model certainly paints a picture with face validity. Let me apply it to the example of buying a new phone. Probably you heard about the phone from someone similar to you, and in turn you will influence people similar to you (such as your colleagues) into buying it. Your mother has quite a different network (you will have only family nodes in common) so it will not be easy to convince her, but it may work. It will be even more difficult to convince a colleague of your mother. Probably the easiest way is to convince your mother first, and she will convince her colleague easily. Well, this is quite a simplistic account, and so it is not likely to be the definite, full and inclusive account of the way influence travels. But the model does show network analysis has something to offer to marketing. Or so it seems. It is a common saying that if the only tool you have is a hammer; everything will look like a nail. Network analysts will only gain insights that can be reduced to properties of the network. In the end we do not know how much information can be found in this way, network analysts are just getting up to speed with uncovering it, and it is hard how valuable this information turns out to be for the practice of marketing. But unless these studies are exceptional, if you are a marketer, birding the network analysis field, seems worth your time. Maybe you’ll spot a penguin.

Reading more.

In my previous blog post called “Modeling the Connected Costumer”, I discussed two other –more foundational – articles from the book the connected costumer.

Possibly it is relevant to look at my post called “Turkles Turn”. There I expressed some concerns about growing amount of networks with a lot of similar people.

The book: The Connected Consumer. Edited by Stefan Wuyts, Marnik G. Dekkimpe, Els Gijsbrechts and Rik Peters is available at Amazon and Google Books.


2 Responses to “The Traveling Influence Problem”

  1. 1 Evaluating the NetGen Argument « @koenvanturnhout MacroBlog
  2. 2 Living Without Money « @koenvanturnhout MacroBlog

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: