Category Archives: Social Media

How To Detect Communities Using Social Network Analysis

Part 2 of our Network Analysis for Market Research series– read part 1 ‘Identifying Influencers with Social Network Analysis’ here.  

Introduction

Social media research isn’t just qual or quant market research translated on to a different dataset – it’s got its own methods. At FACE we’re big believers in using the unique properties of social media data to answer questions that other research methods can’t get at.

And what’s special about social data, particularly on Twitter, is that with sufficiently advanced analysis platforms (Pulsar!) you don’t just collect the message, but also metadata about that message and its author. This provides the information needed to analyse how that message is shared through social networks – or alternatively the network of who follows whom. The result: proper social research that starts from the premise that people are connected, not just atomized individuals.

In the first part of this blog series we introduced some of the possible applications for network analysis in market research, revealing how network data visualization can enable you to identify influencers that have real-world meaning in the context of the social groups in which they belong.

I also discussed how influence exists in different ‘sub networks’ or ‘social groups’, and if we are to truly grasp the structure of these relationships then it’s essential to take these into consideration.  It’s this point that the second blog in the series will expand upon. Here I’ll  look at communities: we all know what these are, but what do they mean in terms of social network analysis? And what can you learn from identifying them?

Why look for communities?

detecting communities

When investigating the role of influencers we highlighted previous research carried out at FACE by Francesco D’Orazio and Jess Owens: the How Stuff Spreads project. In this research we discovered how communities are vital in driving the spread of information. The more communities there are in the audience, the slower viral content spreads, as it takes time to spread between the different groups.

So that’s one reason to understand social media communities – if you’re trying to spread a brand campaign or a piece of content, you need to understand the audiences it travels through. Different groups may well benefit from different messaging specifically targeted to their needs and interests – not one size fits all.

Understanding communities is also important to ensure your influencer program is comprehensive: have you got influencers in all the social groups you want to target?

How are we defining communities?

A community is most often defined as a  group of individuals living in the same geographical location. It can also be used to describe a group of people with a shared characteristic or common interest: the research community, for example Within the social sciences, there is also the approach that views communities as something socially and symbolically constructed, resting on a shared understanding that “I am part of this community alongside these other people”. Political scientist Prof. Benedict Anderson defined the nation state as an “imagined community” (1983).

Using social networks analysis we define communities differently – by looking at how people are connected to each other, and clustering these into similar groups.

So it is a statistical measure of connectedness, and it’s not based directly on whether these people would recognize themselves as being part of the same community. However, what’s so fascinating about networked community detection is that the communities it identifies very often DO have significant real-world meaning, and can help us explore what it is that is defining communities.

How to identify communities? Using a social network analysis program such as Gephi, we can use a clustering algorithm called “modularity” to detect hidden patterns in the network. Modularity looks for groups of people who are more densely connected to each other than would be expected if they were connected by chance. . A network with high modularity has dense connections between nodes within clusters, but sparse connections between nodes in different clusters. As a result all individual nodes (people) in a network can be attributed to a specific cluster, as determined by the modularity algorithm.

A real-world example: my Facebook social graph

Let’s start by revisiting the ego network from my Facebook graph that we investigated in the previous blog. When identifying influencers in the graph I mentioned that it’s vital to pin-point people who the key connectors between different sub-networks on the graph. I was able to provide some real-world context to the data due to my personal knowledge of all the individuals in the network. But even on a small dataset such as this, modularity allows us to develop an even more granular understanding of the relationships.

detecting communities 2

Here nodes are portioned by modularity, with each node belonging to a separate cluster or community, and coloured accordingly. For many of the separate and very distinct clusters on the edges of the network, it shouldn’t come as a surprise that these people belong to their own community.

What is interesting is within the main component, where without the colour coding it’s hard to see any clearly divided partitions. But now we now have four different communities (blue, brown, purple & maroon-ish). So the question is, are these 4 different groups just statistical figments of the network structure? Or do they relate to anything real about the relationships between the people involved?

  • The blue community is made up of people I met at school, all around my age (17% of the network).
  • The brown community is people I went to school with, but also lived close to where I grew up (9% of the network).
  • The maroon community also went to school with me, but all at least a year older that me (7% of the network).
  • The purple community is people I attended college with directly after finishing school (also 17% of the network).

This is a great example of how we can segment individuals by very subtle differences, simply by analyzing the structure of the connections they share.

But how could a network “know” these things about my friends? Well, it’s all based on the connections they have with each other. People who were in the same yeargroup at school are more likely to know each other, and therefore be friends on Facebook – so that’s what connects the real world to the network relationship.

Large scale network analysis

Strictly speaking I could have analysed my Facebook social graph manually – I know who my friends are friends with, after all, so I could have drawn the network manually (though it’d have taken a long time).

But network analysis becomes even more powerful when the analysis is scaled up to a level at which manual analysis is impossible. Using Pulsar to gather our data means we can use network analysis to investigate the relationships in networks of thousands or even millions of people, where obtaining an understanding of the real-world relationships that make up the communities isn’t anywhere near as straightforward.

detecting communities 3

Reverting back to FACE’s previous research into How Videos Go Viral, you can see that modularity and partitioning has been applied on the audiences in the same way it was applied to my Facebook graph. We then applied statistical modelling of the demographics of each group to understand who was in each.

So for the Dove Real Beauty Sketches video (top right), we can see there’s one community averaging 32-year-old white women, in the USA/NYC, working in marketing – and another of teenage girls in Los Angeles who may be white or Hispanic, and who’re into pop music and reality TV. And indeed, it’s that appeal to a diverse audience that made the Dove advert so successful and the most-viewed on YouTube.

How can this work for you?

Think of communities as very similar to the segments identified in a brand’s customer segmentation model. (With demographics analysis layered on, you might even find that they’re the same.)

While direct marketing communications is often customized by segment, historically this hasn’t been something brands have done in social. But, using social network analysis and also Twitter & Facebook ad targeting, it’s possible to send specific messages to specific groups of people.

Powered by Pulsar TRAC these could be people engaging in a specific conversation, individuals sharing a piece of content online, or the followers of an account on Twitter. Any group of people, in essence, as long as we can define that audience through some property of its behaviour in social media – such as keyword, user bio, or location.

Community analysis allows brands to really understand the behavior of their audiences in a way they can’t replicate with offline, non-social data.

It enables brands to get maximum benefit from their influencer outreach and content seeding, by ensuring they’ve got contacts in each sub-community of their audience.

And once communities have been identified, there’s scope for deeper analysis of how each community interacts with brands, the language they use, and the topic . This can allow for truly customized marketing, allowing brands to understand each group’s social media behaviour, and how best to communicate with them.

Network analyses are also great communication tools – each time we put one on screen at a conference, the cameras come out and people start taking photos. We’d love to see more companies going public on their network analysis, and illustrating their audiences back to their followers. As we said earlier, community isn’t just about shared interests but a shared imaginary, a shared recognition that “We are part of the same group.” Sharing community visualisations could be one tool for a brand to create a real “customer community” – moving beyond individualized buyers towards positioning their brand as a source of meaning and identity.

Thanks to Jess Owens for contributing her ideas to this blog post. 

The Samsung vs. Apple court case shows the value of social media research

An excellent case study demonstrating the value of social media research has just emerged from an unlikely source: the Apple vs. Samsung patent dispute.

Apple-Samsung-Trial

Documents shared as part of the court case reveal some fascinating information about how the two companies were thinking about social data in 2013.

It shouldn’t still bear saying in 2014, but the messages seems slow in getting though: social media data isn’t just about “looking back” at campaigns or the last quarter’s KPIs. Samsung recognised the power of social data for “thinking forward”, for understanding customer needs strategically to feed into product innovation and early-stage comms planning. Here at FACE, we think this is an incredibly valuable and under-used use-case.

Here’s how it works:

1. Samsung used social data strategically: to attack Apple

From Neal Ungerleider in FastCo: Networked Insights Reveals How Samsung Used Social Media to Hack the iPhone:

“Samsung took on a company with the arguably most successful consumer product ever created,” Networked Insights CEO Dan Neely told Fast Company. “Samsung asked us how to use analytics to attack Apple.”

[...] Using aggregated online posts and machine learning techniques, Samsung found several specific weak spots where they could outperform Apple. Customers specifically complained about the iPhone’s comparatively poor battery life, the inefficiencies of Apple Maps, how small the screen was, unhappiness with the Lightning cable, the lack of customization, Siri, and the iPhone’s fragility. Samsung felt that it could compete with Apple on most of these points–and, importantly, that they hard data to back up these consumer preferences.

When working with Networked Insights, a big part of Samsung’s strategy was to vacuum up any information on the iPhone 5 that was posted to social media. This meant using the dashboard they licensed to obtain every iPhone-related post on Tumblr, Twitter, Disqus (a popular commenting platform), WordPress, and YouTube, as well as new hits on Google. This information was then classified, as Neely put it, “15,000 different ways.” A big part of the problem for Samsung and others, Neely said, was the difference in extracting relevant information when they needed it versus finding erroneous information on other aspects of individual customers that were irrelevant to the task at hand. That meant a lot of data processing and fine-tuned analytics.

Importantly, Samsung used the dashboard to find what people were posting online about the iPhone–rather than just looking for posts about Samsung’s own products. They then identified specific complaints about the iPhone where their own products outperformed Apple’s products, and tweaked marketing campaigns to emphasize these Samsung strong points.

So: social media research isn’t just about tracking your own brand activity.

It’s incredibly powerful when you search for unmet needs and pain points – what are the gaps where consumer desires aren’t being fulfilled? Do this across a category (e.g. smartphones) or a competitive set (Apple, Samsung, HTC, Sony Xperia, Nexus, Motorola) to identify the “whitespace” opportunities that  aren’t currently being met.

As such, social media has just as much of a forward-looking role to play in innovation and NPD as it does “looking back” at campaign performance and the past quarter’s KPIs. Use it to shape campaigns and communications, not just to measure their impact.

2. Apple thought it was “nuts” to pay for social media monitoring tools. Their loss

Business Insider’s Jay Yarrow spotted something else interesting in the court documents:

Jay Yarow quote

Apple famously don’t do research, you say? No, Apple do do research – but they don’t necessarily do it well, as Tom Ewing recently illustrated.

You’d see the occasional interesting message if you just look at mentions of “iPhone 5″ through Twitter search… But also an awful lot of noise, at a million mentions per day kind of scale. It’d only be through luck that you might stumble across a message that’d spark any strategic consideration.

You want to understand the relative dissatisfaction with battery life, screen size, and poor signal reception? You need a social data research platform. Social media monitoring tools make this data analysable as a whole  in a way that free online tools simply can’t. For example our platform Pulsar (pulsarplatform.com) collects over 1MB metadata around each tweet, making big datasets like this powerfully segmentable by sentiment, channel, hour, influence level, profile bio and other demographics – allowing for a really fine-grained analysis of not just what people are saying, but who and why.

Technology and data augmentations enable the unmet needs to be identified, quantified and ranked. Use a tree graph to visualise the most common words and phrases that follow “I love…” and “I hate…”. Use semantic analysis to aggregate topics, and compare the top topics across the range of positive, negative and neutral sentiment scores. Start coding tweets into clusters, and use machine learning to extend this across the whole dataset.

Through structured analysis, the depth of insight that can be gained from social data is vast – Samsung realised this, Apple didn’t.

3. What we’ve done

This story was met by us at FACE with a nod of recognition – we have been using social data beyond reputation management for many years now.

Here’s a couple of examples of previous work:

i) Mapping the 4G mobile launch

EE Launch Event..Mandatory Credit Tom Oldham/Tom Dymond

Like Network Insights with Samsung, we also dug into what people were saying around 4G to identify complaints and pain points. What topics were driving discussion – signal, pricing, contracts/tariffs, or the iPhone? For each we identified the specific customer pain points our client needed to address in both comms and their product offer.

“WHAT EVEN IS 4G THOUGH I DON’T UNDERSTAND” – tweet, Sept 2013

But it turned out the biggest unmet need was understanding – a high share of discussion came from people expressing their total bewilderment at the new, high-speed mobile spectrum band.  We used social data to identify and categorise people’s questions, helping our client (a mobile operator) recognise and simplify the messages they needed to communicate to help people understand the new proposition.

ii) “Designing Relevance” for Nokia

Here at FACE we’ve been using social data for strategic insight for years. Back in 2010, Francesco D’Orazio and Esther Garland presented at ESOMAR alongside Nokia’s Tom Crawford on how social media research can be used alongside co-creation to produce a better innovation process:

Innovation should not be so much about ‘creation’, but more about ‘emergence’. Defining the boundaries of possible futures means creating the conditions for fostering the emergence of ideas that are already taking shape in the social space, but have not filtered up to the top or are not formed enough to bubble up yet. In a connected real-time ecosystem where the consumer can be as creative as the designer, the new model of innovation should be listening, reducing complexity, decoding the signal from the noise, collaborating with consumers and only then defining the boundaries of possible futures.

The project started with a “download” from social media to gather the widest possible range of themes and scenarios for this project:

The project kicked off with a two week Social Media Monitoring and Trends Analysis programme using netnography, semantic and network analysis across forums, social networks, blogs, news sites, microblogs, video and photo sharing sites from the United States. Using Face’s social media analysis platform Pulsar we tracked more than 100, 000 ‘sources’ (where Twitter counts as one source) and harvested almost 1.5 million items of content. These were analysed to gather insight into how key consumer segments in North America talk about smart-phones and which key themes, topics and angles were most resonant with them. 

Analysing conversations amongst users talking to each other rather than responding to researchers yielded a huge amount of richness. Furthermore, this helped develop clear learnings on language, tone of voice and attitudes to the brand and the category. It allowed for a different kind of research landscape, one which subverts the traditional question and answer format and replaces it with something far more natural and intuitive. By working in a more natural communication mode we also ended up expanding our research agenda to challenges we didn’t even know existed or that we wanted to investigate.

For the full story, read the full whitepaper up on Slideshare here, or check out the presentation:

Or get in touch if you’d like to talk forward-looking social research – I’m at Jessica@Facegroup.com

A Social World of Whisky Part 1: Big Drinkers, Small Talkers?

Winston“The water was not fit to drink. To make it palatable, we had to add whisky. By diligent effort, I learned to like it.” — Winston Churchill

Amongst all spirits, whisky holds a very particular place. From teenagers to world leaders, from whisky and soda to $460,000 bottle – a 1946 Macallan in a Lalique decanter was auctioned at this price in 2010, whisky proves being more than simply a category of alcohol, but a potent landmark of social and economic belonging.

The whisky market is diverse, but can be divided in two main categories: Scotch (i.e. distilled in Scotland and matured for a minimum of three years in oak casks) and non-Scotch whiskies. Both have experienced continuous growth, with some particularly dynamic markets in the last couple of years in emerging countries, especially India and China. Scotch whiskies represent around 85% of Scottish food and drink exports and nearly a quarter of the British total, according to the Scottish Whisky Association.

Such a success in the context of our digital era questions us about the way this phenomenon echoes on social media, how consumers take part into the whisky related social discussion around the world, and what insight can social media bring for the whisky industry.

This blog is the first of a series about the whisky industry that will demonstrate several ways we, as social media researchers, can investigate a broad social dataset and make sense of it thanks to the use of different research techniques and integration of other data sources like sales data.

In this first blog, we’ll have a look at the big picture: identifying how whisky-related social discussion is naturally featuring, and how whisky in social media differs from actual consumer behaviour.

Simply looking at raw social data volumes can be misleading since it doesn’t take in consideration the actual population size of each country, and the proportion of its population using social media. In order to balance the countries’ weight and get a better idea of the countries where whisky discussion is getting more traction, we weighted each country to its population:

Average whisky related social posts per 1000 capita 

Screen Shot 2014-05-11 at 19.30.42

Content posted between August 15th to August 31st,
including “whiskey”, “whisky”, “whiskeys” or “whiskies”.
Collected  by Pulsar, our social media monitoring tool.

What patterns do we see, and why?

Whisk(e)y as a share of British and Irish identity - Ireland is the country eliciting the most social discussion per capita, demonstrating the vitality and weight of the whiskey topic in this country. The second place of United Kingdom in both overall social volumes and discussion per capita, also highlights the importance of the whisky industry and the passion towards this spirit, as home of Scotch whisky – at least for the moment!

The home of Bourbon trails behind Ireland and UK – The United States remains a major country for whisky discussion, especially considering the impressive overall amount of content originating from this territory. But the volumes per capita put this domination in perspective, suggesting that Irish and British are more passionate about whisky.

Whisky proves a healthy topic of discussion in South America and Oceania - A few less populated countries, especially in South America and Oceania, elicit a comparatively high level of whisky conversation, proving their attachment to this beverage, namely Uruguay (6th), New Zealand (7th), Venezuela (8th), and Australia (9th).

Now we’ve drawn a map of social media whisky discussion, getting the most of this landscape implies connecting it to the reality of whisky consumption.

To do so, we are using Euromonitor whisky consumption country data per capita.

Annual whisky consumption/capita (in liters)

Screen Shot 2014-05-11 at 19.35.25

Source : Euromonitor, Worldbank

This data offers us a ranking of the biggest whisky drinkers that we can compare to the ranking of the biggest whisky “talkers”, giving us a new perspective over the whisky market opportunities in terms of social strategy.

Whisky Drinkers versus Whisky Talkers

Screen Shot 2014-05-11 at 19.41.43

* Searches didn’t include words in Hindi, Japanese or Chinese
alphabets, 
so these ranks are likely to be higher in reality

A correlation between whisky consumption and whisky social discussion

Out of the top 10 countries with the higher consumption of whisky per capita, 7 also feature in the top 10 countries with the more whisky related social discussion per capita. However the ranking is quite different…

Less social verbose, more drinking?

Two groups of countries emerge:

On the one hand, countries that feature higher in the consumption ranking than in the social discussion ranking. Including Uruguay, Australia, India or South Africa, this group bears a high potential for social marketers: healthy markets with a lack of social media structure, thus an opportunity for whisky brands to own the category with targeted efforts. The emblem of this group is France, that ranks at the first position for whisky consumption, but only 19th for whisky related social discussion. Some could think that French people drink too much whisky to be able to post their experience on social media. Being well placed to answer this exaggerated statement, I tend to consider that the reason is more likely to lie within cultural and media habits, both in terms of whisky consumption and social media use. This will be the topic of a future blog.

On the other hand, countries that feature higher in the social discussion ranking than in the consumption ranking. And this comprises almost all main whisky producers, namely United Kingdom and Ireland: in addition to a healthy discussion around the whisky consumption itself, distilleries, associations, news websites and organisations contributes to the fact that whisky also feature as a business and economy related topic.
This first glance at the whisky social landscape opens quite a few doors that we will enter in the next couple of months, and that will lead to how we dig more qualitatively into social discussion:

  • Scotch/Bourbon fracture: how is it tangible on social media, and which is winning the social battle?
  • Booze vs Nectar: whisky’s duality
  • A whisky connoisseur social audience
  • The French enigma: understand the specificities of the French social whisky environment
  • Whisky brands: what is their place within the social conversation, and which ones are stealing the show

Stay tuned!

*

anthony

Anthony Fradet is a social media research manager in FACE’s London office. Since gaining a Masters degree from Sorbonne University, Anthony has spent 5 years working for French market research companies, with quantitative, qualitative and social media focus. Before joining Face in 2013, he was responsible for a unique partnership between a top 5  ’traditional’ market research agency (CSA) and a social media research agency (linkfluence). Get in touch with Anthony via LinkedIn or Twitter.

Pulsar update: Visibility 2.0

Today we are introducing a new updated version of the Visibility algorithm that’s powering the Pulsar platform: Visibility 2.0.

The main reason why Pulsar is called Pulsar is that the whole platform is built around the idea of making it easier for anyone to sift through vast amounts of social data by making “important” social media content more “visible”.

One of the key ways Pulsar does this is through its proprietary Visibility algorithm. The algorithm defines “importance” as the ability of a piece of content to reach a larger then average audience and engage a larger than average crowd. The algorithm weights every content on the platform and applies a Visibility score to each post which is then available amongst the metadata used to index and filter the data.
Slide7

Since we launched Pulsar the Visibility Algorithm has been one of the pillars of the platform allowing you to slice any data view (e.g. trends, influencers, topics) by Volume of data or by the Visibility of the content analysed. Below a series of comparative screens that show how different the same social data looks like when analysed by Volumes vs Visibility:

Posts per Day VS Visibility per Day

Slide1

Sentiment Volume per Day VS Sentiment Visibility per Day

Slide2
Top Posts by Volume vs Top Posts by Visibility
Slide6
But the web is an ever-changing ecosystem: new channels are born, new behaviours are introduced, old behaviours evolve to a new scale or disappear and new ways of measuring them are introduced on a weekly basis. In an effort to keep up with the evolution of the web and continue to deliver effective measures of reach and engagement, over the last three months we have been working hard updating the Visibility algorithm.

The new algorithm takes into account:

  • New sources of engagement data, which are now factored in the calculation of reach;
  • New sources of online viewership data which are now factored in the calculation of reach;
  • New sharing and engagement metrics introduced by the new channels we have integrated, such as Tumblr;
  • Raising levels of engagement across all channels resulting in a need for new engagement and reach benchmarks;
  • New behaviours introduced by new channels like Tumblr, where for example the “weight” of a reaction (a re-blog) is completely different from the weight of a reaction on Twitter or Facebook.

Overall, the new algorithm introduces three key improvements:

  1. More accurate audience size estimates for all channels, particularly for News, Blogs, Forums and Review sites;
  2. More accurate engagement figures across all channels;
  3. A more balanced cross-channel view of reach, to enable effective comparisons between the reach of top down and bottom up media (eg. news vs. tweets).

The new visibility weighting applies from April 10 onwards. However, should you want to re-analyse historical data you can extend the reach of the algorithm to historical data from the Data Management interface in the Results View.

We think the new Visibility algorithm is going to help you run better analysis and make more effective decisions and we look forward to hearing your feedback as you start seeing the new data coming through on the Pulsar platform.

If you are not yet using Pulsar and want to know more about Visibility and Pulsar get in touch here.

Identifying Influencers with Social Network Analysis

Part 1 of our Network Analysis for Market Research series by Rob Parkin – read the introduction here.

Introduction

In our work as social media researchers we are regularly answering clients’ questions about online influence and influencers. They know that they’re not the only force influencing perceptions of their brands, and they want to reach out to the other people who are. This could mean identifying the right bloggers to bring on board to increase the likelihood of a successful social campaign, or tracking who is most shaping a discussion about a brand or topic.

Pinning down who is influential isn’t straightforward. The data hardly ever exists to connect a social media message with the actions it may have inspired, such as products purchased or businesses boycotted. Instead what we can really assess is ‘potential to influence’: who’s reaching a big audience, who’s engaging that audience the most and getting a lot of interaction, and who’s demonstrating consistent expertise on a topic. So influence is complex, an outcome of a combination of properties about people, contexts and relationships.

That’s why here at FACE we developed our own proprietary metric to analyse which messages were reaching the biggest audience. Our visibility algorithm assigns each piece of content a visibility score, taking into account the properties of the channel it’s on (e.g. blog content lasts longer than Twitter), the size of the author or website’s audience, and the virality of the post – how many times it’s been shared.

Influencers ranked by Engagement & RTs generated (Pulsar visualisation)

Alongside visibility, we also use Social Network Analysis to understand influence through analyzing the dynamics of online behaviours and relationships. It provides the theory, the algorithms and the software to capture, visualize and explore the data gathered using Pulsar. This can enable us to take influencer analysis to the next level – and it’s what we’re going to discuss in today’s blog.

The role of influencers 

Previous research carried out here at FACE by Francesco D’Orazio and Jess Owens highlighted the role of influencers in how information spreads through social media. It discovered that while influencers may only represent a small percentage of an overall conversation, their role does ultimately shape how information spreads. Tapping into close communities makes content shareable, but top-down influence is essential for content to achieve truly viral speed and scale.

We’ll cover communities in more detail in our next blog, but for the moment let’s understand that influencers play a vital role in shaping conversations, and insight into how their influence is structured can also prove important.

Pulsar_Twitter_Hadfield_Visibility crop for website

Network visualisation of how the Commander Hadfield video was shared on Twitter, with nodes sized by Visibility

Identifying influencers

In essence Network Analysis views relationships as connections. Some people in the network might have only one or two connections (e.g. they only have 1 or 2 Twitter followers), and others might have hundreds or thousands.

So hubs or influencers in networks can be identified by looking for people who are highly connected in comparison to the remainder of the network. Because they’re better connected, these are the people who you may wish to bring on board with an online campaign, to help maximize its chance of successfully reaching the greatest number of people.

So let’s look at an example that demonstrates how networks can help us investigate relationships between nodes and identify influencers.

Investigating my ego network

I’m going to use a very self-centered approach and investigate my Facebook network! I used an application called netvizz to capture the data, and Gephi to perform the analysis.

When compiling a list of influencers you may start with a very basic measure, the number of friends/followers. Using Network Analysis and my social graph, we’ll explore the limitations of this metric, and how we might do a better job.

Introducing my friends & family…..

Rob Identifying influencers 1

In this visualisation the nodes are people who are my friends on Facebook, and the edges are the friend relationships between them. It’s important to note that I’m not on the chart – so the connections aren’t their relationships with me. Instead, the connections shown are the friendships that they have with each other e.g. I’m friends with Amy and Bob, and if Amy and Bob are also friends, there’d be a connection between them. If they’re not friends, no connection.

We can rank nodes by a number of measures; in this instance I’ve chosen degree centrality, which is the number of connections each person has. I’ve used this to determine the size of each node: the larger the node the greater the number of connections. This makes the highly-connected people easier to spot.

We’ve also used what’s called a “force directed layout algorithm” to visualize the graph. This means that linked nodes attract each other and non-linked nodes are pushed apart. So the most-connected people tend to end up towards the middle of the chart.

The first analysis that can be taken from the graph is that a lot of nodes share connections. This why why there is one large giant component in the centre of the graph with lots of highly-connected people all clustered together. This is to be expected as the sample of individuals is taken from my Facebook account, the majority of whom do share common acquaintances.

The thing is, we can also see that the biggest nodes are basically the same size, meaning that they’ve got the same number of connections. This isn’t really telling us the story we need – but using network analysis we can go further.

Identifying Influencers 2

Here we’ve taken the same graph and ranked nodes by betweeness centrality. A betweeness centrality algorithm starts by finding all the shortest paths between any two individuals in the network. It then counts the number of these shortest paths that go through each node. Nodes with high betweeness centrality can be considered information brokers that can connect disparate parts of the network.

The result is a smaller list of potential influencers, pin-pointing the people who are vital in connecting the different sub-networks (i.e. the different social groups) in the wider graph. We have identified four people who are now shown to hold a position of influence on the graph. And the layout of the graph begins to tell us how their spheres of influence are structured.

The person over on the right for example is crucial in connecting two small clusters of individuals to the rest of the graph. I know network analysis has correctly identified this node as an influencer – because she happens to be my girlfriend! So she’s the key person connecting both our families to the larger network of my friends.

How can this work for you?

Admittedly there’s a very short list of people who are interested in the finer details of the network structure of my Facebook graph! Nonetheless it’s an interesting example to demonstrate some of the principles of Social Network Analysis.

What can we take from this example? Using network analysis it is possible to study social groups in-depth, not just as homogenous wholes but understanding them as comprised of dynamic relationships between different individuals. And using data visualization and data exploration it is possible to infer a level of understanding which would be otherwise difficult to get hold of without real-world personal knowledge of the individuals involved.

Using Pulsar TRAC it’s possible to scale this analysis up significantly, sampling mentions by keyword, content or user, and applying network analysis we can powerfully:

  • Identify individual messages driving engagement
  • Explore who is influential in shaping a discussion
  • Map a network of individuals following a brand online
  • Better inform future outreach strategy

Exactly the same methods would apply if we were studying, for example, the community of people talking online about beauty & make-up, or audiophile hi-fi equipment, or photography. We could first find the best-connected people, who a brand might want to target to promote their product to the largest number of people. But we could also find the connectors, the people that allow discussions to travel into new communities and ultimately travel further.

In the next blog in our series we’re going to dive into this further, explore how we can identify communities in network structures and get stuck into some more network analysis previously carried out here at FACE.