Author Archive

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.

Looking for a new challenge this year? Does the idea of working on one of the most exciting and innovative products to emerge in the social media space appeal to you? Well, then this might be for you, so listen up!

We are currently seeking a Product Manager with excellent technical and communication skills to join our interactive Pulsar team in the London office, working on the Pulsar Platform suite of products (PulsarPlatform.com)

Pulsar

Your primary product focus will be Pulsar Flow and Pulsar Live. We’re looking for someone to truly own the development of these products in terms of new features and functionality – managing user needs, defining roadmaps, correcting course and delivering releases on the scheduled date.

You will have at least 3 years’ experience working in a Product Manager or Product Support role, building enterprise-class software products. You will be accustomed to working in a fast paced, technically driven and client facing environment and be well versed in navigating various social media platforms at both a user and technical level. A strong practical knowledge of social media reporting and engagement tools is a great plus.

If this sounds like you, then we would love to hear from you.

Fran-pic.191x110

Visit our Join Us page to download the full job description, then send a CV and covering email to our Chief Innovation Officer, Francesco D’Orazio (francesco.dorazio@facegroup.com).

the FACE Pulsar Team

 

In the first part of this blog series (How Stuff Spreads | How Videos Go Viral part 1: Models of Virality) we identified 7  dimensions that describe and quantify virality. Although none of the variables alone proved able to define a viral phenomenon on their own, they correlated into two models of viral diffusion. One model we called “spike” – the sudden ‘explosion’ of sharing activity – and the other we called “growth”, where popularity is a slower and steadier grower.

Spikers vs Growers

In this blog post, our Chief Innovation Officer and  VP Product Francesco D’Orazio using social network analysis looks at how the audience composition and structure influence the way video spreads.

What makes a viral video spread in one or other of these ways? Most studies on the subject  focus on virality as a feature of the content. But what if virality is (also) a feature of the audience? Can the demographics and the structure of the audience for a video explain how it goes viral?

To recap, we were studying 4 videos:

In this blog post we will show how we analysed the demographics and the social network properties of each video’s audience to understand better how they spread.  Read on for some of our best network graphs yet and some fascinating findings…

Metric #1: Amplification and influence

The first thing we looked at is Amplification. Amplification is a measure of the average “visibility” of the tweets carrying the meme. Tweets with higher visibility imply a more influential status of the author who posted them. Can the influence level of the audience of a meme explain its slower or faster diffusion?

Slide51

Amplificiation is similar across all audiences. It’s fractionally lower for the Turkish protest video and for Ryan Gosling, the first primarily shared in Turkey, the second appealing to a slightly newer (though still very active) Twitter audience. And it’s slightly higher for Dove Real Beauty Sketches and Commander Hadfield. In both cases the variation doesn’t correlate with the virality model of the meme.

Metric #2: How international were the audience?

So the next hypothesis to explain the velocity of the memes was the geographic distribution of the audience. We quantified this as Globality: the percentage of meme shares coming from countries other than the main country. So does the “internationalness” of a video affect its virality? Does a more global or a more local meme spread faster?

Slide52

The answer is again, no. The Turkish protest video was “local” but so was Ryan Gosling – and one spread instantly but the other peaked on day 18. Since both Amplification and Globality seemed not to correlate with one or the other model of virality, we then looked at the demographics engaged with each video.

Metric #3: Demographics.

Does the demographics of the audience affect the way content goes viral? Do young, techie male students from global cities push a meme faster than, say, middle aged housewives from rural Germany?

We used Bayesian statistical inference to analyse the demographics of the audience. This method uses the available information on Twitter and matches it to a sample audience interviewed in real life to get known demographics, across the various countries involved in the study. Below is a summary of the most prominent demographics traits of the four audiences:

Slide54

Slide55

Slide57

Slide56

Although students and global cities feature heavily in all audiences, there doesn’t seem to be any direct correlation between virality models and demographic traits. Instead the demographics are completely different for each meme. Not to mention that students represent 33% of the Ryan Gosling audience, the slowest meme of all – so it seems that youth demographic probably isn’t necessarily a critical cause for a video to go viral quick.

And now some Social Network Analysis…

The audience gets more interesting when you start to look at its social structure. As we couldn’t find any correlation between demographic traits and virality models, we turned to the structure of the audience by mapping the social graph of the people who shared the video.

Your ‘social graph’ is the network of the people you know, and how they’re connected to each other. Because we were studying Twitter sharing of videos, we had easy access to this data through two variables: who each video sharer was following on Twitter,  and who they’re followed by. In technical terms, this gives us a ‘directional’ network with two possible ways for nodes to be connected.  Analysing these connections highlighted some really interesting differences.

Metric #4: Degree, or ‘social connectedness’

First of all we looked at the Average Degree of each audience network. Each person in a network can be assigned a ‘degree’ value: that’s a count of the number of connections they have to other people in the network. We were studying how videos spread in Twitter, so those connections are easy to identify: it’s who they’re following and who they’re followed by.

Interestingly enough, the audiences of the Spiker memes (Commander Hadfield and Turkish protests) are showing the highest levels of interconnectedness – while the audiences of the Grower memes (Dove Real Beauty and Ryan Gosling) show the lowest.

Slide59

Metric #5: Modularity, or “fragmentedness”

The memes that spread faster could do so because the audiences that engaged with them were highly interconnected. But how are this connections organised? To do this we used another social network analysis metric called Modularity. This describes how fragmented the network is and how many sub-communities can be detected based on the density of mutual social connections within clusters of users.

Slide61

The lower the modularity, the less fragmented the audience is into sub-communities, the more cohesive it is and the easier to reach it is. Not surprisingly, the audiences of the Spiker memes are the most cohesive ones, while the audiences of the Grower memes are the most fragmented ones. Cohesiveness and fragmentation becomes much easier to understand when looking at the total number of communities identified within each audience.

Metric #6: the number of Communities

Social network analysis tools allow you to measure the number of ‘communities’ in a social network. Tools such as Gephi provide access to algorithms, such as the Modularity one, that can quantify how people’ s connections tend to gather together into definable ‘clusters’ of closely-connected groups.

Slide63

Slide64

Slide66

Slide65

Whereas the audience of Commander Hadfield is split into 130 communities and the audience of the Turkish protests is split into 51, the audience of Ryan Gosling is split into 382 communities and the audience of Dove Real Beauty Sketches into 1356.

This has a strong impact on the ability of memes to spread through the audience network. Whereas reaching out to just 2 communities is enough to reach 50% of the audience of the Turkish protest, spreading the news to 50% of the audience of Dove Real Beauty Sketches requires reaching out to 8 communities. It follows, then, that where a meme has to travel through more communities to reach people, it moves a little slower – in a ‘grower’ model. By contrast, memes ‘spike’ where they take off in a small number of communities very quickly.

Slide67

So what have we learnt so far?

Yes, the audience’s social structure – the way that people are connected within in – shapes the way something goes viral.

Audiences with a low Average Degree, low connectedness or low density, are more fragmented. The more an audience is fragmented into sub-communities (high modularity of the audience network), the slower a video or piece of content spreads through it . But what causes a higher or lower fragmentation within a specific audience?

Understanding the communities within an audience

To answer this question we tried to measure the demographic diversity of the audiences. The assumption being that an audience showing a higher demographic diversity will also be more fragmented and therefore slower to transmit viral videos.

So we ran the demographics analysis again on the four audiences: this time running it separately on each of the top 5 community clusters identified within each audience. You can see below the results for the top 2 clusters of each audience:

Slide69

The audiences of Ryan Gosling and Dove Real Beauty Sketches show higher demographic diversity, while the audiences of Commander Hadfield and Turkish protests show lower demographic diversity. So high demographic diversity correlates with high modularity and slower meme velocity after all. How is this useful?

How the audience affects How Stuff Spreads

To start with, this means that a meme which is appealing to a broad demographic is probably going to spread slower than a meme that is appealing to a narrow demographic.

This also means that a meme with a broad demographic appeal is going to be more expensive to make go viral. Expensive because it will require more intense paid for seeding/advertising in order to reach out to a higher number of disconnected communities (Dove Real Beauty Sketches is a good example). It may also need persistent replication of the meme to break through the attention of multiple audiences who might not take notice the first time (Ryan Gosling won’t eat his cereal is a good example).

Finally, the organisation of the audience in sub-communities means that influencers lists by subject are pretty useless when trying to reach out to an audience. For example, your top 100 influencers for beauty might well all be part of the same two communities out of the 1356 total communities that make the Dove audience. So identifying gatekeepers and influencers is useful only once the audience you want to reach has been mapped and its communities identified.

The social dynamics of virality

In our previous post we identified and defined two models of virality: Spike vs. Growth.

From this audience and community analysis, we can now augment that with a 3-part model of how content is seeded through groups of people:

1) TRIGGER: A higher than average emotional response to the content triggers an impulse to share

2) VALIDATION: The impulse to share gets then validated against the community the user is part of. This validation happens both in terms of topicality (is this of interest to my audience?) and timing (has anyone else already shared this within my circles?). See this paper for more research on this aspect

3) ESCALATION: The gatekeepers (e.g. media channels, celebrities etc) share the meme helping it reach the tipping point within a specific community. The tipping point is when every member of the community is likely to receive the meme from another member of the community.

Once everyone’s seen the meme and starts to share it on themselves… That’s when you’ve got virality on your hands!

 So what does this mean for you?

Content that generates an emotional reaction is more likely to go viral. People share to say something about themselves. Emotional content helps them figure out easily what it is they are saying about themselves by sharing it.

Picture the audience your content is going to be appealing to, and find them in social media. Learn who they are and what makes them tick.

Your online audience is not a monolith. Online audiences are organised in sub-communities and congregate around key demographics variables such as age, profession, passions and interests.

Map your audience and identify the key communities that are going to help you reach out to at least 50% of your audience.

Once the communities are mapped,  identify the key gatekeepers by community and the connectors between key communities. This will help you reduce the outreach effort.

If your content is going to appeal to a broad demographic expect a longer run and make sure you have the right resources in place for seeding and advertising to a fragmented and  harder to reach audience.

Good luck.

*

Previous posts in this series:

Found this interesting? Got viral content of your own that you want to understand? Check out the tool we used for this study, Pulsar (PulsarPlatform.com) and contact us to arrange a demo – send an email to James.Cuthbertson@Facegroup.com and we’ll be in touch in no time.

Or get in touch with the study authors, Francesco D’Orazio (@abc3d /LinkedIn) and  Jess Owens (@hautepop / LinkedIn).

In an article recently published in Research World Magazine and on his Tumblr blog Abc3d, our Chief Innovation Officer, Francesco D’Orazio outlines the challenges facing the social media monitoring industry – and 10 ways to tackle them. We wanted to make sure to share these thoughts on our blog, as well. 

Social media monitoring is a growing industry but one that is stuck in its old ways. And in need a of a urgent re-think.

Since it emerged 15 years ago, the industry has been largely responsible for driving some of the most interesting evolutions in the research space, such as the democratisation of text mining and computational approaches to mining qualitative information.

This is an approach that for the first time enables both a granular and a birds-eye view of the data, making it possible to produce qualitative observations on a mass scale. A new perspective that is blurring the lines between qualitative and quantitative thinking.

And with computation also comes the ability to mine larger (and messier) datasets, which is in turn steadily shifting the focus of what we call knowledge, from understanding causation to identifying correlations.

However, despite the broad impact of social data on the market research industry and the evolution of its infrastructure, the social media monitoring toolbox hasn’t evolved much in 15 years.

While the web, its users, the brands and the advertising strategies have changed dramatically, most social media monitoring platforms today still do exactly what they used to do when the industry first emerged.

The more than 480 platforms currently available on the market all tend to be affected by similar issues: the obsession with shallow volume-led metrics, the inability to measure exposure, the lack of context to the social data, no understanding of the audience and poor data manipulation and visualization interfaces.

Add to this the more systemic revolutions the industry is facing, such as the visualization of social media, which is going to pose huge challenges to an industry that’s been entirely built on text mining.

Most of these issues can be ascribed to the lack of research thinking in the design of the tools we use today. The majority of the companies that established the frameworks in the monitoring space have come at it from a web analytics perspective. Which has lead to favouring the monitoring and analytics framework rather than the insight and intelligence framework for studying online social interactions.

But the analytics approach has also lead to another big misconception. Social data is not quantitative data, rather qualitative data on a quantitative scale. This might sound like a very byzantine distinction to some but over the past ten years this approach has had huge implications on the way social data has been modelled, analysed, sold, delivered and used by organisations worldwide.

Now with social data intelligence becoming central to many organizations and brands, researchers can and have to play a more active role in shaping the tools for the job. So how can we help disrupting a shallow social media monitoring model to make it more powerful and relevant to the way the web works today?

  1. Introducing new ways of sampling social data beyond keyword tracking: Audience Mapping (harvesting content from a set of users) Content Diffusion (harvesting content that contains specific URLs), Social Simulations (agent based simulation based on social data), MROC Augmentation;
  2. Shifting the focus from the content of the conversations to everything around it: context, behaviors, social graphs and interest graphs;
  3. Implementing solid analysis frameworks to move away from basic analytics towards intelligence, for example embedding techniques to make the most of social data such as social network analysis, discourse analysis, reach analysis, attribution analysis;
  4. Opening up the social silo by connecting social data with other datasets such as sales, NPS, stock trading, media exposure;
  5. Introducing scalable human analysis alongside algorithmic coding by crowdsourcing parts of the research process;
  6. Making machine learning ubiquitous to capitalize on the benefits of human coding;
  7. Improving the Data User Experience to support intuitive data manipulation and delivery at different levels of the company and across multiple devices;
  8. Help re-design the company decision-making process: organizations are learning that they have to re-engineer the way they make decisions in order to make the most of real-time intelligence; there’s no point in delivering real-time intelligence if your client can’t make real-time decisions;
  9. Create smart research products based on integrating traditional methods and social data: live segments, social panels, social surveys;
  10. Making Research Programmable: shape the transition from monitoring platforms to social data driven business applications as social data becomes invisible and embedded into the way any organization creates new products, plans advertising or stocks their stores. Move away from dashboards and Powerpoint reports and plug the data and the intelligence into the products and services we deliver.

Time for the researchers to roll up their sleeves.

*

Found this interesting? Here more from Francesco at our Viral Video Webinar October 23rd when he presents on how different videos go viral differently – and what marketers and brands can learn from it. 

24 hours of global tweets about Sir Alex Ferguson retirement, from the rumour to the announcement to the aftermath through the lens of two visualisation approaches: the streamgraph and the rose.

Streamgraph > Tue 07 May – 10 pm / Wed 08 May 10 pm


Slide2

Nightingale Rose or Coxcomb Diagram > Tue 07 May – 10 pm / Wed 08 May 10 pm

 

Slide1

Data based on 100% of public tweets collected, analysed and visualised with Pulsar TRAC