These presentations delivered by Tim Barker from Datasift and marketing strategist Mat Morrison.
Tim looked at how to use facebook topic data to create customer insight and inform marketing decisions.
Mat looked at the true value of social influence and how to use social data to make marketing decisions.
4. “DATA IN THE 21st Century is like Oil in the 18th Century: an
immensely, untapped valuable asset. Like oil, for those who see
Data’s fundamental value and learn to extract and use it there
will be huge rewards.”
Joris Toonders
Delivered by:
8. 8
Separate Signal from Noise
Derive meaning from text
Extracting Insights while
protecting privacy
Limitless Potential of Insights of
Human Created Data
Fuelled by social, human-created data is
growing exponentially
Big Barriers to extracting
value from data
The challenges of analyzing human-created data
9. Human Data Intelligence
Build a data
ecosystem
Build analytical applications and
insights from a universe of human-
created data
Platform for companies to access,
interpret and analyze human data at scale
12. Surfacing Insights across Facebook
Facebook Page
Topic Data
Posts, Likes and Comments
on brand-owned page globally
Posts, Likes and
Comments on Facebook
13. Challenges in using Social Data for Consumer Research
Data isn’t
representative of
population
Can’t normalize to
remove demographic
bias
Huge effort to interpret
unstructured data
Public networks
have self-promotion
bias
14. Boom! Facebook Topic Data is a Killer App for Research
Data isn’t
representative of
population
Can’t normalize to
remove demographic
bias
Huge effort to interpret
unstructured data
Public networks
have self-promotion
bias
38 Million
People in UK
Surface
Insights
across
Facebook
Insights with
Demographics
Data
Structured for
Easy Analysis
15. Demographics on Posts and Engagement Data
15
Gender Age Range Location
Male
Female
18-24
25-34
35-44
45-54
55-64
65+
Country
State / Region
Self-declared, not
derived.
Break-down the Audience engaging with my Brand by Demographic
Segments
16. Categories and Topics from the Facebook Open Graph
16
CONTENT
TOPICS
Ford (Company)
Automotive (Category)
Going to test drive a new Ford!
Analyse the Topics Associated with a Company, Brand, Event
17. How to start using Topic Data
APPLICATION DEVELOPER ENTERPRISEAGENCY
Topic data enables the ecosystem to innovate
Create a differentiated
social data product.
Leverage an ecosystem
of Facebook-enabled
applications
Build your own data
products. (developers
required)
Access topic data via applications
or agencies.
Create custom solutions that use
insights derived from topic data.
(developers required)
24. THE LAW OF THE FEW
The success of any kind of social
epidemic is heavily dependent on the
involvement of people with a particular
and rare set of social gifts
The Tipping Point (2000)
25. EVERYONE’S AN INFLUENCER
In light of the emphasis placed on
prominent individuals as optimal vehicles
for disseminating information, the
possibility that “ordinary influencers”—
individuals who exert average, or even
less-than-average influence — are
under many circumstances more cost-
effective, is intriguing.
Everyone’s an Influencer: Quantifying
Influence on Twitter (2011)
26. BUT GLADWELL MORE ATTRACTIVE
CELEBRITIES BECOMING CHANNEL
S
“NARROWCAST” BECOMING VIABLE
INCREASED SPEND
TALENT REPRESENTATION
INTEGRATED CAMPAIGNS
38. THE FRY EFFECT
When I tweet a link it usually gets around two
or three thousand requests a second.
Especially if I word it in a way where I really
want people to go to a site.
…
Fifty per cent of the time the site is down in
seconds – even when we've contacted site
owners and they've told us everything will be
fine. It's often an unprecedented amount of
traffic, and they don't have the required
capacity.
“How Stephen Fry takes down entire
websites with a single tweet”
Tech Radar,
March 2010
55. WHAT’S THE VALUE OF AN INFLUENCER?
ACCESS TO AUDIENCEIMPROVED MEDIA EXPOSU
RE
ENHANCED CREDIBILI
TY
INCREASED CONVERSIO
N
IMPROVED SEARCH VISIBILI
TY
ENHANCED CONTENT MARKETIN
G
57. Next up:
ROI & The New Measures of Success
Join us to learn about:
• Using adtech to evaluate campaign performance
• PR measurement in a digital age
Delivered by:
58. Before you go…
1. Say hello to three people you don’t already know
2. Take a photo and share it on social with #techmap
3. Register for a future techmap
Delivered by:
Editor's Notes
Only DataSift is able to provide a “Human Data Platform” that enables companies to Unify all these disparate types of human generated information into a single normalized data model, we then allow to you separate the signal from the noise, enable companies to unlock the meaning from the data by filtering down to just the data they need and adding additional, data classification and scoring and final enabling use of this information everywhere it is needed. This enables DataSift to power an ecosystem of Analytics Companies, Agencies and Brands who can add value to this information and pass along to their customers.
Personal Data Never Leaves Facebook
Just to re-iterate the core of this is a DataSift technology running inside Facebook’s network.
So this whole approach is all about providing insights protecting identity but enabling you to get the analysis and analytics that you and your customers want from that data set.
End of talk track
==========================
Privacy first approach –PYLON has been built from the ground up to ensure privacy, and security to protect users’ private information and identities. PYLON runs securely inside Facebook’s data centers. No personal data ever leaves Facebook. Aggregates of at least 100 people must be represented in a returned dataset, thus ensuring no personally identifiable information leaves Facebook. Data is held for analysis for a 30-day period before being deleted from the service. In addition, analysis can only be performed on content posted by adults. Content from minors is not stored within the service.
We need to make the point that we are the ONLY company that can meet Facebook's privacy standard because we do the processing inside their data centers.
Due to our Privacy first approach personal data will never leave Facebook. This means that the Index must sit inside Facebook’s data center.
This is the main reason we are the ONLY option for companies to understand and analyze this data.
Demographics on posts and engagement data
The first thing is that all of the data that you receive comes with self-declared user demographics. These are the demographics that every Facebook user populates when the sign up. They contain Gender, Age Range and Location. So this is a huge step forward for the social analytics industry. Demographics data is not readily available on any social network. This is because you cannot include it with raw social data as an output. Because we are working in aggregate and anonymized we can add this value to the data that allows you to filter down to a location and see how brand engagement is clustering by different demographic groups. And use that to help inform and shape your marketing and messaging.
End of talk track
==========================
Remember that "Education Status" and "Marital Status" are now "Premium access only"
Categories and Topics from the Facebook Open Graph
The second thing that is really unique here is that there is real-time category and topic detection inside the platform. What that means is that in real-time each post is analyzed to try and identify the topic that is being mentioned or posted about. So here for example on the left hand side you can see a representation of the topics and categories that are trending around the Ford Motor Company. These are the topics that have been identified in the posts about Ford in our index.
And again the thing with this technology is that it really simplifies how you can ensure you have a clean view into the topic you are looking for and any related topics as shown here. An example here on the bottom right hand corner shows how you might use CSDL to gather all of the data for analysis around 3 automotive brands and ensure any content is related to the “Automotive” category.
By utilizing the category you ensure much of the noise will be removed from the terms in your filter and greatly simplifies your overall process for creating filters and understanding topics.
Sentiment Analysis on every Interaction
– each story is assigned a value of "positive", "negative" or "neutral" (there is no indication of the degree of positivity/negativity)
– We are analyzing sentiment for stories in the following languages: English, French, German, Italian, Portuguese, Spanish and Turkish
End of talk track
==========================
7 Ways Sentiment Is Hard To Decipher Online - http://www.informationweek.com/it-leadership/7-ways-sentiment-is-hard-to-decipher-online/d/d-id/1104837?
End of talk track
==========================
There’s a lot of interest in “influencers” these days. Search has trebled in volume.
And the number of articles in the marketing press to meet that search interest (here represented by the BrandRepublic archive) has also increased.
I think it’s worthwhile pointing out that Malcolm Gladwell (we’ll come on to talk about him a bit more) published “The Tipping Point” just around here.
What he couldn’t have predicted, I think, was the explosion in social media marketing that was to emerge a few years later. There were a few blogs around, sure; but MySpace still lay in the future when he wrote the book.
The book more or less set the scene for the rise of blogging, the democratisation of web publishing and social networking.
PR agencies were already beginning to characterise their discipline as the “business of influence”, and the first attempts were being made to industrialise “viral marketing.”
So I’d suggest that Gladwell probably had a profound influence on the way we talk and think about these things. He posited a “Law of the few”, whereby:
The success of any kind of social epidemic is heavily dependent on the involvement of people with a particular and rare set of social gifts.
This is a very attractive narrative, for all sorts of reasons.
But it’s probably wrong. Duncan Watts is a sociologist who’s been studying and publishing papers about social dynamics, advertising and trends since the late 90s. He’s also a bit of a personal nerd crush. He says,
> Gladwell’s law of the few is catnip to marketers and businessmen and community organizers and just about anyone else in the business of shaping or manipulating people. And it’s easy to see why. If you can just find these special people and influence them, their connections and energy and enthusiasm and personality would be put to work for you.
Sadly, this isn’t the case: viral spread has much more to do with the composition of the network than it does with the special powers of any of the people within that network.
> …the most important condition had nothing to do with a few highly influential individuals at all. Rather, it depended on the existence of a critical mass of easily influenced people who influence other easy-to-influence people. When this critical mass existed, even an average individual was capable of triggering a large cascade — just as any spark will suffice to trigger a large forest fire when the conditions are primed for it. Conversely, when the critical mass did not exist, not even the most influential individual could trigger any more than a small cascade.
Sorry, Malcolm.
On the other hand, advertising people have never let the truth stand in the way of a good story.
Thanks to social media platforms, celebrities are becoming channels, controlling and commercialising their audiences. They no longer have to rely on chat shows and press conferences to retail their messages.
And what used to be ‘narrowcast’ platforms (blogs/vodcasts/podcasts) have grown, consolidated, and commercialised. Some of the talent from these are becoming celebrities in their turn.
Advertisers are more likely to want “integrated programmes” that make the most of the talent they employ. That means looking at how they’ll play in social; what additional value they might bring to the table.
And this means that spends are going up. Anecdata suggests that in the US, “influencer marketing” for some clients is a budget line up to 3x media spend.
Increased spend mean that existing talent representation is taking advantage of these trends, and new kinds of talent representation are emerging.
This kind of consolidation and professionalisation makes it easier for busy advertisers to spend money.
Which in turn increases the flow of money.
The only problem is that no-one knows how the hell to value this stuff. There’s no clear, objective way to select talent. We’re all relying on personal relationships, gut feel, and our contacts at talent/artist management companies. It’s all a bit Ari Gold.
If we’re to invest in influencer marketing, we need to know who might we work with if we can’t afford Phillip Schofield. Who could help us reach Asian teens in the North? Who’s hot with Midlands Mum?
It’s pretty clear we’re only going to be valuing influencers in a couple of ways. As channel or as content.
Oprah is channel: she’s the gatekeeper to huge audiences.
George is content. He hasn’t even got a Twitter account. But he commands huge media interest: if we’ve got George on our side, we can parlay that into news coverage, social sharing, organic search traffic and other earned media. He’ll decrease our cost-per-view, and generally improve our media efficiency.
There are all sorts of data streams and cues we might use to assess audience interest in an influencer. Search is a pretty straightforward one: where are we in the influencer’s narrative cycle?
And what’s the relative search demand for each of our influencers?
And of course, what – and where – is the social interest? There are plenty of data to play with here.
So now let’s look at the channel side of the story.
It’s frequently remarked that there are things that are worth measuring and things that can be measured; but that not everything that can be measured is worth measuring and not everything that’s worth measuring can be measured.
Nowhere is this more true than when it comes to social media. We spend an awful lot of time measuring what can be measured, without ever asking whether we should be.
Twitter’s sheer volume of data, and the relative ease with which those data may be collected often mean that we don’t look any further than that. Most of the tools for identifying and evaluating “social media influence” simply use Twitter data.
I’m a big one for tradition, so I’d like to start with Twitter.
Stephen Fry is a great example. He’s pretty big. So big, in fact, that when he tweets, he can send two or three thousand clicks a second and bring websites to their knees.
Which is pretty awesome, really. If this is true, he’s a one-man DoS attack.
Can we validate these numbers?
Every so often, Mr Fry tweets a link that’s been shortened with Bitly. You may be aware that Bitly provides public stats for those links.
So, I went and collected his last few thousand tweets, found that about 10% of them contained bitly links, then counted the clicks on each of those links. This is a histogram of that data. I do enjoy histograms.
From this analysis, I can tell Stephen Fry gets a median 1.6K clicks-per-tweet.
This number is impressive, but it’s hardly the 3,000-4,000 per second that were reported.
Now, I’d like to be fair to Stephen Fry, and point out that he probably only has website owners’ words for the magnitude of the Fry Effect. I mean, how would he know how many clicks he’s sending?
The lesson here is, don’t trust other people’s data. You don’t know where they found them.
So let’s hypothesise that traffic isn’t a great proxy for influence. Surely Fry’s 9.5 million followers mean something in terms of exposure? That’s an audience that competes favourably with X Factor or Britain’s Got Talent.
Only we don’t know how many of Fry’s 9.5m followers see each tweet. It would be naive to believe that each tweet reaches all of them.
Twitter doesn’t publish reach figures (although individual accounts can see their own impression data.) How might I begin to investigate?
Fortunately, a generous friend of mine shared his Twitter stats with me. He has around 100,000 followers; but on average he nets about 5,000 impressions with each tweet.
His impressions set a top limit on his potential reach, so let’s be optimistically sloppy, and say he’s reaching an audience equivalent to about 5% of his followers.
I’m also interested in retweets-per-tweet as a metric. It strikes me that this is a good way to tell how effective a given Twitter user is at activating their audience.
We also have lots of convincing evidence that reach (or at least the impressions that we’re using as a good proxy for reach) increase with retweets.
My friend receives a median 4 retweets-per-tweet1 (or 1 retweet for every 25,000 followers.
Fry gets a massive 113 retweets per follower. But that only works out at 1 retweet for every 85,000 followers; suggesting that my friend is more than three times as good at activating his audience.
There’s a reason for this. For a long time Fry was (and may even still be) one of the “Suggestions” for new Twitter users to follow. Twitter doesn’t want new users to start with an empty timeline, so it encourages them to follow popular accounts. If you were looking for such a thing, it’s a great example of the Preferential Attachment.
Paradoxically, the “Suggested Users” decreases the overall quality of a users’ followers. New users are much more likely to churn (leave Twitter, never to return) than active users of the service. So it’s fair to say that really famous Twitter users have a certain amount of inflation in their numbers. It’s not their fault: it’s Twitter’s. So let’s look at another platform.
For all sorts of reasons, YouTube is more exciting than Twitter for advertisers. Making short videos is well understood by the business. There’s a strong paid media platform. And we’ve seen an explosion of young, talented creators who’ve established close, direct relationships with their loyal audiences.
Money is beginning to flow in: certain categories (beauty, lifestyle, food) are better represented and funded than others; but as those niches fill, other content areas are strengthening.
Oddly, there aren’t many really great data sources for YouTube analysis (please do tell me if you know of anything.) Instead, YouTubers are often ranked (and valued) according to their subscriber counts. Let’s take a look at that.
Here’s a typical YouTuber, with a typical price: £15,000 to have them create a video around our brand. They’ve got 1,200,000 subscribers, so a naive calculation puts that at £12.50 CPM (for non-advertising people, that’s “Cost per Mille”, or cost per thousand views.)
Video CPMs typically range between £8 and £25, so this feels like a bargain.
So I counted the views on every video this YouTuber had made. Each dot on this chart shows the views for one video. There are a few stand-out amazing films; but the majority are clustered near the bottom.
Another way of looking at the same data (I did tell you I like histograms). You can see there’s a pretty tight grouping between about 50,000 and 150,000 views per video. This YouTuber gets a median 111,000 views per video: or about 10% of his subscriber count.
So instead of netting out at £12.50 per video…
…we might actually be paying £135 for those “organic” views. That’s more than 10x the naive estimate.
Another thing: demographics aren’t public. You need to ask the YouTuber or their agent for them. Often, though, you’ll see something like this: YouTubers’ audiences can skew young and female. And maybe not in your own country.
This sort of thing isn’t restricted to YouTube: the price you pay with earned media on the web is a certain lack of control over targeting: you don’t get to say who you reach unless you invest in paid media.
This really is only a series of somewhat related thoughts. But to summarise briefly:
Talking about “Influencer Marketing” isn’t helpful. No-one shares a good definition of “influencer” and very smart people who I trust call into question their very existence.
I’d strongly caution against adopting new metrics or KPIs for influencer marketing. Instead, try to judge influencer activity against your existing activity. Does adding incremental “influencer” activity increase efficiencies at an acceptable cost?
And I’d strongly recommend using paid media to amplify and target content that you create with the influencers – we’ve seen excellent responses to promoted retweets, for example. On the whole, though (and for all the reasons cited) I instinctively prefer to think of talent as co-creators rather than as channel.