At a time when attention is a scarce commodity, true power lies in understanding how information flows within networked audiences. It is no longer possible to demand one's attention, or even expect it at a certain point in time. For a message to spread, it must be picked out from overflowing streams of updates, photos and links, and chosen to be reposted by each individual. The networked nature of social media may give some messages an overwhelming boost in popularity, but in most cases they fade as fast as they were created. It is imperative that we use available data to better model, track and gain insight about our audiences in order to make the best decision at any given time.
43. Know your audience
What type of people follow you, how they’re connected, where in the world they
are
Alignment between content and audience
No to vanity metrics
more followers doesn’t necessarily mean more engagement
Realtime Information
be aware of the constantly changing conversation
Bridges vs. “Influencers”
just because we’re connected doesn’t mean information will flow
Building up the Right Network
trust, resonance, interest
Not necessarily about being First
My name is Gilad Lotan and I'm the VP of Research and Development at a New York City startup called SocialFlow. SocialFlow is an optimization service, helping media companies, brands and effectively anyone with a networked digital audience make better decisions about what content to publish and when - say the right thing at the right time. I'm here to talk to you about my work on digital audiences with a focus on network analysis and information flows. I look at social networks, and model an understanding on how and why information flows. What gets certain pieces of data to propagate, while the majority dies very fast.
There are a number of ways which we attempt to answer these issues. I will not go into the tool for lack of time. But what we do is:Rank attentionResonance PredictionGraphing engagement over timeRealtime TrendsTrends over time
In an environment where the threshold to publishing and consuming content nears zero, attention has become the bottleneck. One cannot demand attention anymore, or expect to have it at certain times of the day. We all need to understand the preferences and behavior of our respective audiences, and adapt our own behavior in order to attract the attention of users.
information flows through people. Networks of people who decide at any given point in time what interests them, and what they choose to repost to their subsequent networks. In order for messages to propagate, people along the way must pay attention: notice them at the right time and pass them onwards. Data helps us paint a picture of our audiences, in effect see their digital silhouettes. - what do they talk about? - when are they active? - what do they say when they reference me? - who else to they care about? - how are they interconnected?The more data we have, the better we see, the better decisions we can make when talking to invisible audiences.As we do that, the better we can asses the value of a piece of content given an audience.
We're used to static models: given a TV channel + show that's on + the demographic information about the likely audience for that show, there was a dollar value that would be given to ads that would run during the commercial breaks. The value for this TV spot is based on historic information collected over very long periods of time. The value we assign to the estimated level of attention that a piece of content will receive is based on models built over time, mostly based on historical data.
Within social network spaces, the equation is very different. In this 140 character attention economy, true value lies in understanding how information flows - how people manage what they choose to give attention to. We're not broadcasting anymore, and cannot asses "value" of targeted group just based on static information. Things are dynamic and networked, with qualities such as: topicality, timing, influence and trust heavily affecting the spread of messages.
In this presentation I’d like to …Highlight the power of dataLike looking left/right before crossing the streetTrue power in making sense of data
In order to asses dynamics of an event as it happens, we need to know what the baseline regular patterns look like. This way we can track whats out of the ordinary, and get a better sense for how unique or important an event might be. Effectively giving us the ability to normalize or data - standardize our calculations across the board, so that we can better compare results. Important so that we can capture deviations from the normDiurnal patterns – Lo and behold: Humanity isfairly consistent!
Obvious: if I’m targeting a US audience I’d behave different than when targeting an Indonesian crowd!
We model what’s normal
When we dive into that spike, we can see a discreet pattern emerge. Sports games tend to look exactly the same (+- differences in volume). Initial excitement is represented as the smaller spike to the left. And excitement builds up towards the end of the game (massive spike). By mining these types of shapes over time, we have the ability to identify an event as it is happening without knowing in advance – for example music concerts, holidays and breaking news.Certain events draw certain types of curves, certain levels of participation.
People come together around events. In this case:MarchaAntiEPN – in blue – weekend protest that brought 90k people out to the streets across MexicoTonyawards – note the spikes that happen every time an award is announcedTrue blood – fan excitement buildup towards the TV screening.
Systems are partially algorithmic. Algorithms aren’t always aligned with people’s expectations and/or perceptions
Our media ecosystem is more and more governed by AlgorithmsTwitter’s trending topic algorithm
Once we identify events, is dive into the actual data shared during the event. Mapping them helps us get a sense for the conversation that’s emerging.Hashtags have emerged as a way for people to gather around topics or events. Way to map our what’s central around a topic – and shifting relationships between topicsLots of snarkWay for candidates to interact with fans/followers
Sheds a light on people’s perception of the candidates.- Mitt romney: #gayrights, #lgbt, #jesus, #flipflop, #jobs, #economy- Newt Gingrich: #palestine, #OWS, #immigration, #abortion (he famously said – “Stop whining, take a bath and get a job!”Equal: #republican, #dems, #economics, #amnestyWe see this data in realtime, and how it changes over time…Co-occurence
Now that we have a better sense for the data, lets talk a bit about how we quantify audience engagementHistoric notions of audiences come from a broadcast and performative perspective, where there exists a shared experience of viewership: people collectively focused on a screen or performance. With the advent of the internet and services like Twitter and Facebook, traditional media is expected to engage with their audiences, shifting from the one-to-many broadcast means of communications to a networked model. This completely changes the dynamics, requiring entities that have been strictly controlling of few communications channels, to navigate multiple streams, targeting each for a different audience type. Constantly tuning your media/messaging based on your audience – looking at how well your posts are doing w/r/t your audience
Social offers us a constantly evolving, dynamic interest graph in realtime. Next we look at what ppl from an audience are talking about. This is Al Jazeera’s audience. Maps out hundreds of thousands of conversations happening at a given time. Using force directed algo – organize graph by co-occurrence – gives us a topical mapping, snapshot, of whats currently central to an audience, what they’re attentive to.Social Media – networks self organize around topics
Differences between audiences- All these methods as a way to understand which audience is more “engaged” and in what way
Or they can come together around media outlets:In this case we clearly see the diurnal pattern which suggests a substantially higher level of usage around the US timezones. But we also see that Al-Jazeera’s audience has a much wider curve, due to the geographic distribution of its audience. Additionally we see all audiences spike when breaking news occurs.
We can see how it evolves over time – different clusters emerge as the day advances
We can also see how they’re interconnected
Giving a face to our audience: 1) topics 2) connectionsPre-existing clusters within the audience – can assume where they are from
Key points: Looking at data different ways teaches us different things about what it represents By adding social graph data, we can extract important details about the group of followers
Media Outlets gained great value from breaking the news: much more attention focused on them. Now its rare that news has not leaked out to Twitter first.How does this change the ecosystem?
There was much speculation on why the presidential announcement had to take place on sunday night. Some were on the Gaddafi side, and others, Bin Laden. Interesting about this story – You wouldn’t consider keithurbahn an influencer based on his profile – not many followers
Media Outlets gained great value from breaking the news: much more attention focused on them. Now its rare that news has not leaked out to Twitter first.How does this change the ecosystem?
What you see here is the node representing Keith Urbahn along with all of the retweets he generated, along with Brian Stelter from the NYTimes, and the responses he generated. Before May 1st, not even the smartest of machine learning algorithms could have predicted Keith Urbahn's likelihood to spread information on this topic, or his potential to spark an incredibly viral information flow. While politicos "in the know" certainly knew him or of him, his previous interactions and size and nature of his social graph did little to reflect his potential to generate thousands of people's willingness to trust within a matter of minutes.
Not necessarily about being first – But needing the right kind of network!In this case – the information was fed to the network, but nobody listened – no-one propagated until the big guns came in