Skip to content

Social Networks and Consumer Targeting

2012 May 23

Who is your target consumer?  That’s been the “go to” question of branding consultants for generations. People who come up with insightful answers are valued and admired.

There has been no shortage of strategies and methods put forth to answer it.  Consumer surveys, focus groups and statistical techniques have sought to identify, communicate to and acquire the right consumer for the right product.

Yet it’s becoming clear that we are, in large part, a product of our social networks. Our opinions and behavior are greatly affected by a complex web of social influence. That realization is leading to a new mindset and a host of new strategic possibilities.  Network analysis has been useful in other fields and the potential for marketing is enormous.

Who Led the 9-11 Attacks?

Within a day of the 9-11 attacks, the leadership structure of the hijackers was published in major newspapers.  How did they figure that out?  If you think about it for a minute, that’s the ultimate targeting problem and not all that different from identifying who would be a likely consumer for a brand.

The answer is, of course, secret and classified, but network scientist Valdis Krebs performed his own social network analysis by piecing together links between the hijackers from news reports and calculating three measures of influence:

Degree Centrality:  How many links each hijacker had to the rest of the network.

Betweenness Centrality:  Their location in the network relative to other members.

Closeness Centrality:  The average social distance between a particular member and all of the other members of the network.

This is what he came up with:

You’ll notice that Mohamed Atta led by a wide margin in all three measures, so it’s not surprising that he was identified as  the leader of the network.  It’s also interesting that Zacarias Moussoui, the reputed “20th hijacker,” figures in so prominently.

For a more complete picture, check out Valdis Krebs’ full summary here.  It’s a fascinating read.

The Hunt  For Bin Laden

Of course, after the initial attacks, there was the matter of locating the main orchestrator, Osama bin Laden, an effort depicted in Peter Bergen’s new book Manhunt.  In 2005, after years of fruitless search, a paper called “Inroads” began to circulate among intelligence officials that outlined 4 “pillars” to form a “grid for the search:

Al Qaeda Leadership: The most obvious place to look for the world’s top terrorist is to study his consorts closely.  Some we captured and interrogated, others we found later and scoured their files.  Unfortunately, we never found much to lead us to our target.

Family:  No man is an island.  Even arch-criminals have people who love them.  Other top fugitives, most notably Adolf Eichmann and Pablo Escobar were tripped up that way. Bin Laden had a large family that he was devoted to, so this was an obvious place to look.

Media Statements: One of bin Laden’s trademarks was his high profile media statements.  By tracing back from the destination (high profile media outlets) investigators hoped to locate their origin – the man himself.

Courier Network: As the leader of a large organization, bin Laden had to communicate with his followers.  Officials knew that he had long since stopped using any kind of electronic form of correspondence, so they surmised that he must be using a network of couriers. This was indeed how they eventually found him.

When I read this, a few things struck me.  First, it sounds a lot like the way we segment brand consumers, but using social groups.  Second, it wasn’t the high profile connections that bore fruit, but the lowly courier that led to the target, which is exactly the opposite of what you would expect to see in a marketing presentation.

The Influentials Myth

It’s the last point that’s particularly important.  We marketers tend to assume that the world is relatively straightforward.  We break consumers up into neat little segments and expect them to act accordingly.  Rich people should buy expensive products, poor people should pinch pennies and so on.

Nevertheless, the real world has its own logic that isn’t so easy captured on a PowerPoint slide.  Nowhere is this as true as with the myth of influentials, seemingly magical people who are so cool, hip and powerful that we all dance to their tune.  In fact, there is very little evidence that influentials exist and much that says they don’t.

In reality, influence is more of a function of thresholds than anything else.  We are not convinced by anyone one person or event, but by a tangled web of multiple signals.  The overall effect can be substantial and research by Christakis and Fowler has shown that many behavioral attributes, including obesity, are affected by our social networks.

Just as in the case of bin Laden’s courier, anybody in the network can tip the balance. After all, networks are a group phenomenon and looking at individuals, to a large extent misses the point.

The Power of “Q”

If we want to think seriously about networks and marketing, we need to get away from the old thinking of a “target consumer” and start thinking about what a viable target network would look like.  In other words, what type of network will carry our message efficiently?

A highly cited study done by Brian Uzzi and Jarrett Spiro gives us a tool to determine exactly that.  They examined the social networks of Broadway musicals and measured a metric called “Q,” which, without going into too much detail, basically measures network density.

Here’s what they found:

When people with no previous ties start working together, the results tend to be poor, they’re not familiar enough with each other to develop a strong working process.  Teams who have some familiarity with each other fare much better.  As network density increases, so does productivity.

However, past a certain point they get to know each other too well (i.e. the network becomes overly dense) and creativity, along with financial performance, suffers. Networks that are too loose don’t allow information to flow, but ones that are too tight can keep new ideas out and squelch innovation.

Notably, the type of social network was more determinant of a play’s success than any other factor, including marketing budget, production budget and past performance of the Director.

Brands as Networks

Much like Broadway musicals, brands are also networks.  Successful ones are dense enough that consumers influence each other, but not so inbred that they are closed to the outside world and reject change (a trap that can make growth difficult for brands with cult-like followings).

So, beyond the $100 billion Facebook IPO last week, we should see the emergence of electronic social networks as an opportunity to look at consumers in a whole new way that can allow us to ask entirely new questions.

Did that last campaign improve our network by adding to our consumer network or dilute it?  Should we try to improve network density or reach?  Social network analysis can reveal answers and take us in completely new directions.

Of course, none of this dispels the need for more conventional analysis and tactics.  To be sure, there are considerable technical and computational hurdles for managing network metrics for large brands, but network science will have a lot to offer in the years to come.

– Greg

6 Responses leave one →
  1. May 23, 2012

    Excellent post Greg. In my research, betweenness centrality is usually the most revealing of the centrality measures. In networks that measure any nd of flow, it’s a great measure of who is really critical. Eigenvector centrality is pretty useful too, though it’s not used as widely for some reason. Although, it’s sort of snuck in the back door as people have started to use PageRank, which is a very similar measure.

    The point about brands as networks is very interesting – I think there is definitely something of substance there that’s worth further thought.

    [Reply]

    Greg Reply:

    Thanks Tim!

    btw. I thought eigenvector centrality and closeness centrality were the same thing (i.e. eigenvector is a methodology for calculating closeness). Is there a difference?

    – Greg

    [Reply]

    Tim Kastelle Reply:

    The short answer is that they’re similar but different. The long answer is more than I can write on my iPad right now. But yeah, not completely the same…

    [Reply]

    Greg Reply:

    Okay. Thx:-)

    – Greg

  2. May 24, 2012

    OK, now I’m at a keyboard so here’s the deal on centrality measures.

    Degree centrality is exactly proportional to degree, so I don’t know why anyone uses it. But tons of people do. If you just talk about it as number of connections, that’s fine, but when you add centrality to the title, many people act like you’re talking about something different, which you’re not.

    Betweenness centrality takes all of the shortest paths between every possible pair of nodes in the network, and measures the percentage of these paths that a particular node is on. It’s hugely useful, and like I said, in my experience, one of the most important network measures to look at.

    Closeness centrality is the average of the path lengths from each node to all the other nodes in the network. A smaller number means that you can get to all the other nodes more easily from the focal node. It seems like it should be a pretty useful measure, but it’s not used that widely. (see this: http://www.sonivis.org/wiki/index.php/Closeness_Centrality )

    Eigenvector centrality is actually more like degree centrality than closeness. It combines the number of connections you have with the number of connections your connections have (PageRank is basically an eigenvector centrality measure). So if you have a small number of connections, but they’re all to hubs, your EC will be higher than someone who has many connections to only poorly connected nodes. EC is a pain to calculate, but very useful as well.

    Mark Newman explains these all really clearly in his paper The Mathematics of Networks, which is well worth reading if you haven’t yet. It has the equations for everything in it too, which freaks some people out, but I think his writing is very clear.

    [Reply]

    Greg Reply:

    Thanks Tim, that’s extremely useful. I’ve been meaning to get around to reading Newman’s paper, but never have. Thanls for the push.

    If anybody else wants to read it, you can find it here: http://dl.acm.org/citation.cfm?id=1809753

    – Greg

    [Reply]

Leave a Reply

Note: You can use basic XHTML in your comments. Your email address will never be published.

Subscribe to this comment feed via RSS