Twitter Friends and the Influence of Influentials in Word of Mouth Marketing


Social Networks that Matter

Without going into links to specific posts, I’ve noticed a trend among many blogs I try to keep up with over the past couple of years. I can’t count the number of times I’ve seen prominent bloggers post publicly about having to pare down the list of RSS feeds they read, or tweets they respond to. Since Peter Kim’s blog is the most recent instance of the trend I’ll use one of his recent posts as an example of what I mean. Peter noted that he increasingly hears an echo chamber across social media blogs in which the same content, case studies, anecdotes, etc. gets repeatedly posted and commented on. More cynical observers might contend that the complaints about information overload from influentials is a little like strutting in front of a crowd. Nevertheless, it is difficult to dispute the point that attention is a scarce resource on the Web. So is engagement.

Ross Mayfield recently pointed to a study published by researchers at the Social Computing Lab of HP Laboratories that addresses the point succinctly by pointing to constraints on friendship in directed social networks such as Twitter. A directed social network is characterized by an absence of explicit reciprocity constraints, fifty people can follow one person without that person necessarily following any of them. First Monday’s most recent issue includes an article, Social Networks that Matter: Twitter under a Microscope, that reports on a study of Twitter users by Bernardo A. Huberman, Daniel M. Romero, and Fang Wu of HP Laboratories.

The authors analyzed data from 309,740 people using Twitter. They compared the network of interactions people actually engage in while using social computing technologies such as Twitter to the network of connections with whom one shares a social relationship. Networks of actual interaction are considered networks that matter by the authors.

By networks that matter we mean those networks that are made out of the pattern of interactions that people have with their friends or acquaintances, rather than constructed from a list of all the contacts they may decide to declare.

In other words, the research focused on reciprocity as well as connection in studying the social network of Twitter. 

Hidden Friends, Influentials, and Dunbar Numbers

JP Rangaswami over at Confused of Calcutta offered the observation that the HP study’s findings probably ought not surprise us. The researchers distinguished between three types of people using Twitter: followers, followees, and friends. Anyone who uses Twitter is familiar with the first two categories. We choose to follow the tweets of some people on Twitter and others (not necessarily the people we choose to follow) follow us. However, the research also looks at tweets where hidden friends send @messages to specific people. Hidden friends are defined as people who received at least two @messages from others. The authors report that “90 percent of a user’s friends reciprocate attention by being friends of the user as well.”

Rangaswami contends the HP study is significant because the data presented in the research seems to call into question the assumption that social software works to raise Dunbar’s number. To quote from Wikipedia,

Dunbar’s number is a theoretical cognitive limit to the number of people with whom one can maintain stable social relationships. These are relationships in which an individual knows who each person is, and how each person relates to every other person. Proponents assert that numbers larger than this generally require more restricted rules, laws, and enforced norms to maintain a stable, cohesive group. No precise value has been proposed for Dunbar’s number, but a commonly cited approximation is 150.

The point is an important consideration for anyone thinking about the use of social media in word of mouth, or viral, marketing. Rangaswami points out “that as the number of friends increases, there is apparently no loss in reciprocity.” Indeed, the HP authors note that “reciprocity of attention is a very consistent trend as it holds for both users with many friends as well as for users with very few friends,” and plays a crucial role in defining the ‘hidden network’.”


However, Rangaswami adds a crucial point. He observes that the data from the HP study (Figure 4 — see below) indicate that Dunbar’s number, or some variation of that number, constrains the scale of networks that matter. And Rangaswami  proceeds to note, “there is a suggestion that the number of friends is constrained in Dunbar-like manner.”


Rangaswami  continues to think that social software can shift the Dunbar-like constraints currently evident in online networks.  Ross Mayfield, on the other hand, thinks that social software helps people manage their Dunbar number more effectively.

What Social Software can do for your Dunbar number is help what goes in and out of the 150.  Help you discover people you should add to your network.  Stay in touch with old friends so they can be active friends for at least a moment in time.  Help others sift out of view.

Indeed, Peter Kim’s solution to the echo chamber he sensed in his RSS feeds involved dropping connections and adding some new ones.

Twitter and Word of Mouth Marketing

The authors of the HP study contend that one of the most important points to take away from the research relates to the pattern of interactions in the social networks supported by Twitter. 

Many people… see online social networks as an opportunity to study the propagation of ideas, the formation of social bonds and viral marketing, among others. This view should be tempered by our findings that a link between any two people does not necessarily imply an interaction between them. As we showed in the case of Twitter, most of the links declared within Twitter were meaningless from an interaction point of view. Thus the need to find the hidden social network; the one that matters when trying to rely on word of mouth to spread an idea, a belief, or a trend.

My take on the findings from the HP study is as follows: the influence of influentials on word of mouth, at least in Twitter, increases as members of their hidden network reciprocate their own followers through interaction, rather than increasing due to the total number of people who follow any particular influential. Placing an emphasis on the latter rather than the former increases the echo some hear in social media.

To me, one of the key issues from these findings is whether they speak to the big seed/little seed issue in relation to viral marketing. Anyone reading this post care to share their opinion?

Share this post…

add to : Add to Blinkslist : add to furl : Digg it : Stumble It! : add to simpy : seed the vine : : : TailRank : post to facebook

Posted by Larry R. Irons

13 Responses to Twitter Friends and the Influence of Influentials in Word of Mouth Marketing

  1. bbbozzz says:

    Larry, a fascinating post – have you taken a detailed look at the HP methodology? I personally would be very be interested in your thoughts.

  2. Larry Irons says:

    bbbozzz (or is it John?)…thanks for reading Skilful Minds. No, I have only seen what is described in the article, though the same publication is also available from HP’s site…

    From what I can tell, the methodology in the HP study is limited in one key way…Twitter does not provide a separate status for followers and friends, at least currently…therefore, the HP group distinguished “friend” status from “follower” by those people using @user messaging (direct) rather than indirect messaging to the network of followers…the arbitrary definition of a friend as someone receiving two or more @user messages seemed limiting to me until I read the study of Twitter usage in Akshay Java’s dissertation from this past December (MINING SOCIAL MEDIA COMMUNITIES AND CONTENT)…Apparently, Twitter used to allow for explicit distinctions between “friends” and “followers” and Java’s research was done on a dataset during that time period…Interestingly, the HP study found around 25% of all posts were directed by @user messaging, whereas Java’s data found 21% of users posting @user messages…so these two recently studies of Twitter seem close…I wasn’t using Twitter in 2007, and I’m still getting my feet wet using it…

    Java notes, “When we collected our data [2007], Twitter’s social network included two types of directed links between people: friend and follower. A Twitter user can “follow” another user, which results in their receiving notifications of public posts as they aremade. Designating a twitter user as a friend also results in receiving post notifications, but indicates a closer relationship. The directed nature of both relations means that they can be one-way or reciprocated. The original motivation for having two relationships was privacy – a microblogger could specify the some posts were to be visible only to her friends and not to her (mere) followers. After the data was collected, Twitter changed its framework and eliminated the distinction, resulting in a single, directed relationship, follow, and a different mechanism for controlling who is notified about what posts.”

  3. Tim Tracey says:

    “By networks … those networks that are made out of the pattern of interactions that people have with their friends or acquaintances, rather than constructed from a list of all the contacts they may decide to declare.”

    Absolutely! An effective approach to social networking will, ironically, be a near perfect copy of “real” person-to-person, social networks. This approach will bring the web-based social network to the “real” WOM network.

    Why not simply “webinize” the WOM process that has consistently delivered the most qualified leads? Why can’t businesses increase their sales by connecting to their network of satisfied customers, neighbors and friends in a Web-based social network like Facebook?

    Two requirements for such a network are spelled-out at

    Such a system would eliminate the time and expense of creating and updating a Web page. A simple, user-friendly design would be as easy to set up and use as a Facebook page, while allowing businesses to know exactly where their best new customers are coming from. (They could even voluntarily reward them to show their appreciation.)

    As we say at, “Reward the community by empowering trusted relationships.”

    – – Tim

  4. Larry Irons says:

    Good points Tim…

  5. Adrian Chan says:


    Great post. i think there’s little doubt that in talk tools like twitter, which are time-based and conversational (of a form), the Dunbar number, while constant, probably includes a smaller number of active conversation participants.

    Let’s say that some percentage of the Dunbar number is a close set of friends, with whom daily interaction is not necessary to sustain engagement and maintain the relationship — but with whom that conversation might be very grounding, rewarding, and meaningful.

    There might be another percentage that is a set of peers — members of one’s network with whom coded and informative exchanges serve to surface, explore, share discoveries and create collaborations.

    And there might be some percentage given over to new contacts, or more accurately, twitter partners in talk — transient network members with whom a relationship is latent but not yet enduring. People for whom we are available for talk, but with whom we have no explicit commitment to maintain contact. The conversational activity among members of this subset would be more governed by the etiquette and practices common to the social tool in use: twitter is not blog commenting is not facebook friending is not linkedin answering and so on.

    I would like to see some research into twitter networks that is diachronic — which tracks conversation over time and correlates that with follower/following count.

    I would expect that the number of transient relationships increases with an increase in followers/following. Does the Dunbar number hold steady? Or is it the wrong metric altogether for conversation monitoring? I suspect it’s the wrong metric. Our ability to sustain engagements would more likely be a matter of our attention spent on the site/service, our interest in it (which goes through phases), our “goals,” our experience to date and historically with the site (rising interest after adoption, plateau, fade out, rediscovery….), and of course the runs of talk themselves (talk increases around cultural news and events).

    I would imagine that these conversation engagement metrics would also correlate to user personality types, and to the differences between monological, dialogical, and relational (Self, Other, Relational activity-oriented) “archetypes” of people in general.

    To wit, a Self-oriented person might talk more if s/he believes he commands a bigger and more attentive audience. Stats revealing traffic to his site, click throughs on his links, retweets and @replies will embolden his/her engagement and make him/her more enthusiastic about tweeting.

    An Other-oriented person might talk more the more @names and Directs s/he receives. Being inclined to respond to people, and to engage in one-to-one conversations, this user’s increasing following count will likely create more conversations — but possibly very passing and transient ones — as many of them are of course greetings and introductions (what we do when we meet people).

    A Relational/activity oriented person might @name @name @name people more the more s/he sees group activity on twitter. This being the kind of interaction that is least well supported in twitter (multiple D messaging isn’t possible, for example, cutting out backchannel chat). Chat-style communication, which is necessary to create a sense of communal or group involvement and interaction, isn’t possible in twitter. So the relational/activity oriented user must sustain an awareness of social groups over time — this is a gate to group interactions. [I’m finding that Yammer, which I use with adhocnium members, is a twitter-chat tool for me. There’s no sense that a public reads our posts, and we conduct a slow chat over Yammer that in which, almost paradoxically, the @reply becomes a sidechannel!]

    A smart marketing tool would thus not use influence, but would use conversation dynamics and transient properties of social media conversations and their participants, to determine not who to impress, but rather how to distribute by means of user-centric social media communication networks.

    I’ll put this in Benjamin’s language: Communication in the age of its technical mediation is contingent no longer on the interaction handling of facework but on the loosely-coupled coordination of asynchronously sustained individual commitments. I nearly called them “commentments.” (reference is The Work of Art in the Age of Mechanical Reproduction – Walter Benjamin)

    (This became so long that I’ll blog it on my site, too. Thanks for the inspiration — keep it going!)

  6. Larry Irons says:

    Hi Adrian,

    Enjoyed your post/comment. As always, your thinking is both interesting and stimulating in its attention to what is social about social media. Not enough people who use social media to inform strategy ask that question. I don’t disagree with your point that the Dunbar Number is limited as a way to explain the range of conversational dynamics over time. My point, following that of Ross Mayfield, was that social software such as Twitter allows us to manage those dynamics. I take that to be the point of your quote from Benjamin whose work I’ve read and appreciated. Am I wrong?

    As an aside, my personal favorite from Benjamin is his work on Baudelaire.

    The HP study essentially makes a point about reciprocity in directed social networks. My take on their findings, aside from the question of the Dunbar Number’s relevance, is that influence does not depend on the number of followers or friends of an individual twitter user. Connections, as you rightly note, are not engagement. Distinguishing the influence of people by the number of followers, or the number of people they follow, seems like the equivalent of traditional mass media’s concern with audience to me. Although this is probably a trivial, taken-for-granted assumption by informed users of social media, I can’t tell you how many ads I’ve seen lately for social media consultants in which the number of followers of a candidate is treated as an indication of how much social media savvy they possess.

    If we look beyond Twitter to the larger category of using social software in marketing, I agree with your comment that,

    “A smart marketing tool would thus not use influence, but would use conversation dynamics and transient properties of social media conversations and their participants, to determine not who to impress, but rather how to distribute by means of user-centric social media communication networks.”

    As far as your archetype distinction goes, I think talking about archetypes or personalities is treating the medium as psychological rather than social. People come to social media as individuals and, as you rightly note, the responses they experience shape their overall pattern of engagement. However, I’m more inclined to think their patterns of engagement don’t result from personality as much as their overall grasp of the communication afforded by the application architecture and the responses they experience over time as they participate in it by communicating with others.

    Your recent tweets indicate an appreciation for the dynamics of gift giving in social networks. Gouldner provided one of the best updates I know of to the classic literature of Durkheim, Mauss, and Simmel on the topic. He distinguished between the norm of reciprocity and the norm of beneficence to explain the dynamics of how gift giving kicks off relationships of exchange that reciprocity maintains. Unlike traditional, or industrial, society, gifting on social networks is almost entirely symbolic though that doesn’t mean the reciprocal connections made remain symbolic.

  7. […] report from researchers at the Social Computing Lab of HP Laboratories that I read snippets from yesterday makes the point that a link online does not mean an […]

  8. Black Lotus says:

    Black Lotus MTG…

    […]Twitter Friends and the Influence of Influentials in Word of Mouth Marketing « Skilful Minds[…]…

  9. estrategias marketing…

    […]Twitter Friends and the Influence of Influentials in Word of Mouth Marketing « Skilful Minds[…]…

  10. Thanks a bunch for sharing this with all people you actually recognize what you’re talking about! Bookmarked. Please also consult with my website =). We could have a hyperlink trade agreement between us

  11. I think the admin of this site is truly working hard in favor of his website, since here every data is quality based data.

    • Larry Irons says:

      Thanks for the compliment. It is difficult to keep up with comments given the amount of Spam. Wasn’t sure about your contribution so I gave it the benefit of the doubt.

Leave a Reply to Tim Tracey Cancel reply

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

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

Facebook photo

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

Connecting to %s

%d bloggers like this: