Social network analysis and PR

For those who I haven’t bored to death talking about it, one of the things I’ve been reading up lately is social network analysis (SNA). I’m not claiming for one second to be an expert and probably opening the door for me to be shot down. But here goes…

Fingers crossed its going to solve all the problems of the PR industry, if not I’m wasted approximately 100 hours of my life. At the same time though, there’s nothing really all that special about it, like Social Object Theory, its just another way of thinking about communications.

From what I’ve read so far, social network analysis concerns itself more with the flow of information and the great thing is that it is all measurable (yep, numbers and that). No doubt, there will be some of you reading this post thinking “but its about the content, its about the conversation” – which is entirely true, but let’s stop being daft – we are not going to start sending people turds on the stick because we have a new toy to play with. This is just a way of targeting to make sure the message reaches the right people.

If you want an explanation looky here: – I’ve spent too many nights writing and re-writing the introduction of blog posts, trying to explain this stuff in a simple way, as such I now can’t be arsed. This post is going to dive straight in.

Defining influence
At Porter Novelli, we’ve been using our own network mapping tool to help identify ‘influential’ people in a network. ‘Influence’ is a bugger to define – is it popularity? reach? a combination of both? (if so how much emphasis would you place on popularity vs reach? 50/50? 70/30?) or is it something else entirely different? (See Jonny Bentwood and the gang at Edelman‘s crack at analysing the influence of Tweeters in their Twitter Index to look at just how complicated this stuff can be).

Within SNA, there are lots of ways you can measure the importance of a person/blog/website within a network. Below I’ve highlighted a few of them.

Look! A diagram!
To illustrate my points I will be using a basic sociograph called Krackhardt’s Kite – a popular reference point in SNA. It’s a simple undirected graph, which although there’s slightly more to it, basically means that if I read your blog, you automatically read mine. However, the principles are largely the same with directed graphs. I’ll try and explain things as I go on (I’m trying to write this so that you guys know what I’m talking about, but any proper SNA geeks stumbling to the blog don’t think I’m an idiot). If you need any of the terms explaining, this site is a good reference point:

Degree Centrality
This is roughly translated as popularity. Quite simply, it is the number of lines (sometimes called edges/arcs depending on whether they are undirected or directed and also known as paths, semi-paths, walks, etc, depending on what you are looking for – can you see why this blog post has taken me so long to write?!) the little round yellow things (the proper term is nodes, vertices or actors, again depending on what the network is illustrating) has. I think the correct definition of degree centrality is the number of ‘neighbours’ a node has. In the diagram above, node 7 has the highest degree centrality. It has six other nodes linking to it.

In the blogosphere (Hello Geeks!), this could be the number of inbound links a blog has. In the UK PR industry, I’m guessing node 7 would probably be your Neville Hobson‘s,David Brain‘s or Geordie chav, Stephen Davies‘ blogs.

You would target these when you want to break a story to as many people as possible very quickly. You might not reach everyone but it would create a massive, immediate impact. For example, if you had an exclusive story that was breaking, these guys would be the ones you would target.

It’s no surprise to find that the blogs I’ve literally plucked from the air have been going for some time. ‘Preferential attachment’ describes how popular websites become even more popular as time passes. If they are already an established, influential blog, then new bloggers will continue to link to them in the future. With reference to communications, PRs should be looking to develop these long-term relationships where possible in the same way they cultivate their journalist relations.

In ‘traditional’ PR, nodes with a high degree centrality would probably be the mainstream media.

Closeness centrality
This is a measurement of how close one node is to all the others in the network. For example, for node 7 to pass a message to node 9 it would have to pass through three lines. In the example network, nodes 4 and 5 have the highest closeness centrality. Mat suggested in a recent blog post:

Instead of looking for WOM influencers, why don’t we look for areas of high potential — and target those people who are likely to be
receiving lots of WOM stimuli?

Nodes with a high closeness centrality are more likely to be to be exposed to a message in a network. Conversely, if we targeted a blog with high closeness centrality, then the information would be spread throughout the network much easier.

But let’s look at it another way. This is not strictly closeness centrality (I’m just making it up as I go along…) but related. What if we singled out a specific node which spread the information to the whole network much more quickly? For example, if we seeded information with node 5 it would take 4 steps (including the initial seeding) for the message to spread to the entire network.

However, if we seeded content with node 8 we would see this:

Therefore we would target node 8. Apparently, I’ve just read that this is actually the Markov centrality, but I haven’t done enough research to really comment on it, so instead, happy to admit I’m making this stuff up.

With regards to PR and blogger outreach, I can see this targeting being more relevant when focusing on smaller, possibly niche groups – in the SNA world also known as cliques. Let’s say we wanted to target a group of IT directors based in lovely Bradford who blog (by my reckoning, there’s about seven of them – there’s only 100 computers in the whole of Bradford believe it or not) with a consistent flow of news. You have budget to build a relationship with them but Node 7, the most popular guy doesn’t have time to go for lunch (he works also works night shifts as a taxi driver and Sundays in the local chippie). Who would you target to build a relationship with? Possibly 8? If you need to make a major last minute announcement (e.g. the venue of an event has moved), targeting node 8 would get the message out the quickest.

Betweeness centrality
This is the one I’ve been using when describing SNA and how it works in comms. It basically sounds most impressive and ‘new’. According to Exploratory Social Network Analysis with Pajek (don’t ask…), Betweeness centrality rests on the notion that “a person is more central if he or she is more important as an intermediary in the communication network”. These nodes when taken out of the network restricts the flow of information (in the cases where they stop the information from flowing between two groups they are known as cut vertices).

Lets take for example two offices based in New York and London, the bosses of which both have PAs. Top level information flow between the two offices would more than likely involve the two PAs, neither are particularly popular (in the degree sense, not how nice they are). They would be the ones with high betweeness, take one of them out and the flow of information stops.

Another way of looking at it is by introducing a concept called ‘The Strength of Weak Ties‘. Without going into too much detail, it explains how your good mates might be good for a laugh down the pub or a visit to the strippers (it doesn’t use this example in the paper), but they are unlikely to tell you something you didn’t already know – their friends are your friends already. A weak tie, on the other hand, is someone you are acquainted with. Because they do not necessarily share the same friendship group, they are actually more useful because they can introduce you to new people or pass on new information.

The world needs these intermediaries to facilitate the spread of information and share ideas.

In the first diagram, the node with the highest betweeness centrality would be node 8. In a massive consumer campaign targeted at lots of different segments (and with limited budgets) these bloggers would be the ones who you can count on to spread the message between different sets of people.

Another way to look at it is during a crisis – a blogger has some how managed to get their hands on some highly confidential information and you don’t want it to break into the mass media. He’s decided to blog about it but only has about five readers who all only have a handful of readers each anyway even if they decided to blog about it. The exception is that one of the bloggers’ readers is known to be a well known journalist. It is this person that you should be most concerned about contacting to broker some sort of deal.

Much more to this stuff than I’ve written about

With my limited PR experience, I’ve tried to illustrate the strengths of each with examples with how you can look at influence in different ways. There are others such as Eigenvector centrality (which is not about how many people read blogger A’s blogs, but who reads it – Google uses this as a basis for its search engines), random walk centrality, bridging centrality, etc which I haven’t even touched upon yet. The truth is you would be thinking about all of them (and to be honest they overlap a lot anyway). Although I don’t know for certain (I’m looking into it now), I suspect you would place more emphasis depending on the type of network you are investigating. It’s possible to work out just how degree/closeness/betweeness centralised a network actually is and base your targeting around that.

But this is about efficiency. Although many PROs will claim to be experts at building relationships with bloggers, the truth of the matter is that there are just too many. You might have had an occasional pint with Wadds or Brucie, but how often will you be contacting them to pitch stuff (unless you just happen to be a plant seller and stumbled upon Wadds’s recent unhealthy infatuation with allotments). We work within an agency and probably work across loads of different sectors, I haven’t the time to really build relationships with many (and don’t lie, you don’t either), SNA could be a way of picking out the most effective ones.

Then it depends on the type of information you have. In an informal group, I’m likely to tell my friends about my workmate Roger‘s heroic quest to stop drinking for the whole year, despite my teasing. However, my friends would not tell their friends about Roger. Jesus this is getting complicated now. Competitions, news, reviews – they are all different types of content. Therefore, would you change the story (like you would/should with journos) depending on a blogger’s position in a network?

What about the campaign? Long term? Short term? Consumer? B2B? Does it change again?

There’s also an issue that I’m assuming the information will always flow from A to B via the shortest path (the correct term is its geodisic distance). Often this is not the case and while I might subscribe directly to Mashable, I might read the news on Jed or Jaz‘s blog first.

Are we thinking too much about this? I had a pub conversation with Marshall Manson just over a year ago when we were talking about metrics. Marshall said that you know it’s right in your gut. Though he’s got about a million years more experience in PR than I have so that’s easy for him to say.

It would be great to get old school PROs thoughts on the three criteria I’ve outlined above. I can seriously see SNA changing the face of PR, I just don’t quite have the experience of some of yous yet.

Right that’s taken me the whole of Saturday. See I told you, I’m still alive. See you in six month’s time.


41 responses to “Social network analysis and PR

  1. Interesting ideas, I particularly like the counterintuitive concept that the supernode isn’t necessarily the one to target. The biggest challenge is finding enough time to get interns to trawl the data for you.

    One thing I would say however is remember the limitations of a model; otherwise we could make as much a cock-up of marketing communications as economists have of the economy.

  2. Pingback: Ged Carroll – backtype | CockBot

  3. Really good piece mate. Some familiar ground, but I love the diagrams and the idea of the “strength of weak ties”. A lot of food for thought.

  4. In a rush to do a tender today, but will definitely return to analyse this great post in more detail. The point about not targeting the supernode is one that I’m often talking about.

  5. Good post fella. If it makes PRs engage with bloggers better then they do journos, then some of these ideas must be a good thing. One issue, I guess, is what happens when they start targeting the same node – then maybe a weaker node might be a better option..?

  6. nice work, Timmy H.

    Interesting timing – I’m interviewing someone this week for New Media Knowledge who sells a product to manage SMA. Not sure quite how ‘automated’ that can be, but I’ll let you know when the story’s up.


  7. @stuart – I totally agree with you there Stuart. Often, the bigger the node (certainly with the influencer analysis anyway), the more protected and less likely to respond it is.

    Is the trick therefore, to try and agitate the periphery to create the groundswell the chief node can’t ignore?

  8. Very impressive post that explains this stuff clearly and gets me thinking. Keep em coming.

  9. Nice one Tim.

    I don’t think its long enough though. Slacker.

  10. Nice chunky, detailed post Tim. Thanks for sharing. Hope you still saw some daylight on Saturday 🙂

  11. Tim – this looks like it represents six months’ worth of thinking, and it’ll take me a good long time to really digest all you’re saying. But since you and I talked about closeness centrality a couple of weeks ago, I *can* add what I was going to tell you about my investigations on that matter.

    You know this already, but it’s worth pointing out to others: the kite network you’ve drawn is essentially undirected (all links point both ways.) Examples of undirected networks include Facebook and LinkedIn – while examples of directed networks include the blogosphere and the Twittersphere.

    I think that this can cloud our thinking on this matter.

    Anyway, thank God for Instapaper.

  12. Interesting post.

    Another factor to bear in mind I think is that journalists and big bloggers often like ‘discovering’ a story on a smaller blog, with the benefits that then brings in terms of willingness to write about in a prominent way.

  13. All good stuff! Antony Mayfield is the Don when it comes to networks.

    Check out this e-book

  14. The idea of the “strength of weak ties” is really interesting – thanks for this… really got the grey matter going.

    However “I can seriously see SNA changing the face of PR” is seriously over-egging it. You are indeed over-thinking this.

    We PRs need to distinguish ourselves from the plums of advertising planning – one way to achieve this, in my opinion, is to minimise jargon and fight PowerPoint-itis… I’m afraid parts of your article reminded me of those PowerPoint charts the US Army used to plan the Iraq war… (

    Thanks for the article though, I enjoyed it – keep em coming!

  15. Tim this post is excellent – as I mentioned on Twitter I’ve been working on a response to Mat’s post from a couple of weeks ago and now I’ve got to push it back again to incorporate this one. 🙂

    Tim (and others above) who are interested in “strength of weak ties” would definitely enjoy a book called “Getting A Job” ( by Mark Grannovetter, which is all about network theory and might* have been the text that originally set forward the notion.

    *It’s been a while since university so forgive if I’m incorrect there.

  16. @everyone – thanks for your responses. As i admitted throughout, I haven’t got the experience in PR to see whether this works or not. It’s just a really interesting way of thinking about communication flow. What I’ve read so far on SNA, it strips much of the context from the process of communication. I quite like that though and if it makes us take a step from saying our role is to generate buzz, conversations, etc, then it is no bad thing. The trick is putting it all together.

    @ged agreed on the issue of an over reliance on models. My thinking lends itself to models because I’m never THAT confident in what I’m saying, and models sometimes help me justify my actions. I was speaking to a senior PR person the other day who still talks about the fact that PR should own ‘the conversation’. While I do not disagree, I think we need something in place to provide better evidence as to how we are ‘owning the conversation’ (if indeed we are…).

    Also what’s your pingback about? (cockbot!?)

    @davidbrain agreed, there’s definitely nothing new – just a culmination of thoughts. And thanks for the mention on Twitter.

    @stuartbruce looking forward to your thoughts. RE: Supernode – I’m definitely not saying we dismiss them outright, just that there are other, just as perfectly valid options.

    @theredrocket cheers for stopping by Phil. I don’t think it is an issue if people start targeting the same blogger. It depends on the blogger, how receptive they are to the message, how much time they have etc. Journalists may be contacted very regularly but they are use to it and their job is to respond. On the other hand a blogger (with a weak position) may not be targeted as much but then he probably writes one post every 6 months and doesn’t really like being pitched is one example.

    @ChrisLee hey buddy, nice logo. There’s sections of this that can be automated, such as gathering the data (it’s just 0’s and 1’s). But I’m not for one minute suggesting that the targeting process is automated. It’s still making human decision, just with more information at your disposal.

    @paulfabretti good point about the groundswell, and I think your point underlines the fact that a story still isn’t a story until the big boys (the mass media) pick up on it. For your first point see my comment to @redrocket.

    @udayradia where’ve you been!

    @jordan Can you see why i said it’ll take me ages to write this post before. Absolutely epic.

    @philsheard – it was a mate’s leaving party on Sat night and thought after spending the day working I’ll get to the venue early and have a cheeky pint on my own. He didn’t get there for another 2 hours – suffice to say I got drunk on my own. You decide whether that’s a good or bad thing…

    @matmorrison Instapaper is ace! That’s where i read about the Markov centrality (when I was drinking on my own –see above – and pretending I was texting people so I didn’t look sad). Agreed wholeheartedly that directed networks change things somewhat – I thought it would overcomplicate things on my first post about SNA. But if we were to have the information for directed networks, I think UCInet and Pajek would still be able to work out the centrality measures related specifically to the direction of the flow of communication.

    The only thing that concerns me at the minute is that most things I’ve read (certainly the more old school books) suggest that it doesn’t matter about which direction information flows (certainly when measuring betweeness I’ve heard this several times). That’s wrong though surely? We are trying to measure influence therefore the node sending out the information is more influential than the nodes receiving it? I’m confused…

    @Markpack excellent point. Again I’m stripping out context here, but it’s how Eigenvector centrality works. Basically if one blog is read by only a handful of people but one of them happens to be Barack Obama then the blog automatically becomes more influential.

    @MarkHanson will take a look. Mayfield’s been quiet since the NMK gig though hasn’t he?

    @BenCaspersz OK, probably overegging it and already admitted that I think too much about this. I just think that its a way of thinking about communications in a completely new light. I’m certainly not saying I’m the first to think about it like this (basically everything I’ve said, Mat’s told me about at some point). I think SNA will go some way to helping us crack the measurement debate. It won’t solve it at all, but will give the PR industry more proof points that we can ‘own’ conversations.

    Again sorry about the length of it the article. As Mat rightly points out it is 6 months hardcore reading and thinking, braindumped onto a blog post – I could have easily done this in short steps, but I was scared that I wouldn’t write another post for another couple of months or be bored by SNA by then. My diagrams are awesome though 

    @Fernando Mark Grannovetter’s a legend. Everything I’ve read references him. Looking forward to your response.

  17. Nice post Tim – I am impressed.

  18. @chrisnorton cheers for stopping by. beers (no shots please) next time?

  19. Thanks for the link @Mark Hanson…

    @Tim – Quiet? I suppose so – divided between work and writing a book in my spare time.

    I did a WOMMA webinar on social web literacy last week at:

    …& I’m still saying things at:



    Anyway, yes networks and understanding them are absolutely key to understanding how the web works. Hence Brands in Networks, the title of the e-book.

    It’s important to keep coming back to network theory to challenge how you are doing things and shake yourself out of lazy, channel-centric thought habits.

    Once you’ve worked your way through books like Linked and Six Degrees for the basics, I recommend people take a look at Beinhocker’s The Origin of Wealth, which looks at financial markets through the lenses of complexity theory and evolution.

    Take that analogy into the idea of the super-complex networks of information and relationships that are the web and you have some interesting approaches to how communications and influence work online.

    Lovely stuff – look forward to reading more of your thoughts on this. It’s all about the networks…

  20. Pingback: renaissance chambara alias Ged Carroll - Links of the day

  21. @Amayfield

    What book are you writing? I’m hoping that you are writing the script for the sequel to Maid in Manhattan, desperate to know what happens next.

    The thing I like about the networks analysis is that it strips out content (you’d be daft to ignore content altogether) but it almost purifies the process of communication in a way.

    Thanks for the book recommendations. I will try those other books, I think I got carried away and chose the hardest ones first.

  22. Yes, Tim, it’s basically Maid in Manhattan 2. In Dorking,

    It purifies how you look at it, but

    A mathematics professor who specialises in networks told me that there is only so far you can go with network analysis models before they become so complex that they are basically meaningless.

    The problem is when they are simple enough to understand they mask a lot of complexity.

    So you end up with models. But they are just models – they help test, calibrate, guide analysis and decisions. But the truth about the way networks work in the real world, whether it is diseases or press releases moving through them, is always more complex – and therefore impure, unpredictable, erratic… as well as surprising and delightful.

    I guess what I am saying is that understanding the principles of networks will help you to be a better communicator. Being “network literate” as Howard Rheingold says, will help you be a better 21st Century Citizen. But I don’t think we’ll ever have the full picture, as in being able to plan and predict outcomes precisely, of how communications and influence in networks works. Nice as that would be…

  23. @amayfield Agreed definitely, I like the purity because it allows you to think clearly (and i like the pretty pictures).

    The whole thing about SNA is that for me, it’s a case of having more information and insight in our decision making. Its about putting together network analysis, with content, with relationships, with context. Then hoping that you get lucky.

  24. My concern is that all the theory can get so complicated and resource-sapping that it ends up crowding out other important planning considerations – like finding out hard facts, listening to gut instinct and using common sense – and campaigns end up being over-clever, airy-fairy and ultimately falling on their arse…

    Planners have a tendency to hide behind their complicated PowerPoint slides.

    If I sound a bit cynical it’s because too often it’s been me who’s had to stand up in front of the Board for the final evaluation and the person who came up with the complicated theory – which no one can quite remember – has evaporated and left some other chump (yours truly) waffling round it and looking like a numpty.

    ; )

  25. @Ben

    I think there needs to be a balance between the qualitative and quantitative, definitely. I don’t want this post to turn into a conversation vs conversion debate because we’ll just go around in circles like every PR debate inevitably does. – just food for thought and a response to the couple of people have been asking why I’ve been quiet lately.

    It’s easy to hide behind diagrams because its easy to blag. Hopefully, I’ll be able to explain this stuff in a simple way (let’s be honest – I’m hardly a genius – I’ve a vocab of about 16 words).

    Numpty (!) haven’t heard that in ages 🙂

  26. Enjoyed your article. Would be good to some more examples in action.

    You might want to check in with Dr. Walter Carl in Massachusetts his work on WOM is interesting, and very relevant to your article.

  27. Thanks John,

    I’ll take a look though he appears to have stopped blogging from April onwards?

    and thanks for following on Twitter (though I don’t actually say anything useful on there)

  28. Check out Walter’s company, he also has a great list of WOM articles.

  29. Pingback: Oops. Reading list | Rock Star PR

  30. Tim, you got your step counts wrong, but like you said… you were just making this up. 😉

    We often look at how nodes are embedded in their network neighborhoods before we know the flow of influence or exposure. Finding the influentials has more to do with their surroundings than with them.

    Good Luck!

  31. Hi Valdis

    Good point about neighbourhoods. Will be following up on the post soon.

    Step counts – I’m assuming you don’t count the first one? (it didn’t feel right including it, but just looked better in diagrams…)

  32. Great article Tim. I nice follow up would be about how to use SNA software (any will do), which seems to be another minefield

  33. Pingback: Diffusion in PR « Much ado about nowt

  34. Pingback: Twitter analysis of @LabourList followers « Just Juiced

  35. Pingback: Just Juiced

  36. Pingback: Labour’s Twitter Tsars | Just Juiced

  37. Pingback: The 14 most essential tools for a killer Twitter strategy - Gabba. Create Conversation.

  38. perfeito!!!!parabens!!!!

  39. Pingback: Social network analysis and PR | Much ado about nowt | Infromation Blog | Free SEO Tips and Resources |

  40. Very awesome post. A very good read.

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