WORKSHOP AGENDA
Start: 09.00am
The social data challenge
– Chris Shaw, CEO Commetric
Data smart social enterprise
– Maz Nadjm, Founder SoMazi
Break out challenge
– 3 Teams
Summary & close
Close: 11.00am
3
TOPICS
1. Gathering data:
– What to look for, what to ignore and how to get
intelligence, not just information, from social
technology
2. Internal stakeholders:
– How to build bridges within your organization
with social insights
3. Social impact:
– How to report social media impact to elevate
your profile with key decisions makers
4
SOUND FAMILIAR?
Social Measurement - speaking one language in the company
it would be good to
understand how to
optimise SEO through
social data findings.
6
YOU’RE NOT ALONE
Q. What is preventing your
organisation/clients from harnessing social
data as effectively?
7
WHY IT IS EXTREMELY DIFFICULT
Mega trend: Avinash Kaushik, industry guru
talks about the ‘fragmentation of data
capture’.
• Different apps give rise to different metrics making
holistic view challenging
– New behaviours such as Pinterest ‘pins’
– Acceleration of new functions on social networks to retain traffic, notably
Facebook
• Organizations simultaneously under investing
– 63% of organizations have1-2 people tasked with social data &
measurement (Altimeter)
8
BUT WHY IT IS IMPORTANT
Q. What have been the primary positive
impacts of social media data within your
organization?
9
FUNDING LOOKING MORE LIKELY
Q. Where does digital business rank on your
strategic agenda in 2013?
10
CURRENT PAIN POINT PARADIGM
Comms CRM
Comms
Structured
Self ServiceEarned KPIs Data
KPIs
Dashboard Data
Open
Single Networks
Campaign
Text / Tweets
Accurate
Single Format
Qualitative Platform
Chinese Latin
languages languages
Multi Format
Platform Quantitative
Pictures Indicative
/ Videos
Closed Continuous
Networks Programme
Executive
Unstructure
Sales
Business Owned Report
Business
d data Data KPIs
KPIs 11
11
ONE APPROACH: TECHNOLOGY + PEOPLE
We combine advanced text
processing and visualization
technologies to optimize
human
analysis.
Harvesting multiple media types for
360° perspective
Accurate Natural Language Processing
(NLP) for speed
Statistical tools combining text with
business performance data
Visualizations to reveal new
connections in data
12
SOCIAL DATA SCORECARDS
The key is to develop KPIs that align to the
business objective yet deliver relevant
insights to key stakeholders
Business objective
Paid Earned Owned
Editorially Socially
Controlled Generated
KPIs for key Stakeholders
KPI DATA FRAMEWORK
Challenge is to combine data types and
analysis methodologies to deliver accurate
KPIs
HARVESTING QUANT FILTERS QUAL FILTERS
Editorially Socially Unstructured Data Verification &
ControlledGenerated Analysis
Structured Data
KPI SCORECARD OUTPUT
Combining quant & qual methodologies,
editorially controlled & socially generated
media formats into stakeholder KPI
scorecards Quant
Total Buzz vs. Buzz by Network
Buzz by geog market
Owned vs. shared coverage
Sentiment (indicative) by network
Qual
Discovered messages by brand (post / ref content)
Coverage drivers by peak
Sentiment by brand/competitor
SOCIAL INFLUENCE DATA FRAMEWORK
Deliver auto rankings of indicative influence.
Will improve but currently requires
qualitative enrichment
QUANTITATIVE FILTERS QUALITIATIVE FILTERS
Inputs Auto Rankings Relevancy x Consistency
- Overall activity
- Posts / topic
- No. followers
- Topic authority
- Re-tweet ratios
- FB ave. likes /post
- FB ave. comments
/ post
SOCIAL INFLUENCE DATA OUTPUT
Our objective is to identify ‘consistently
passionate influencers and map ‘gatekeepers’
to wider networks
ARAB SPRING – YEMEN
SOCIAL DATA / BUSINESS DATA
What are the data types that can be used to
discover the business impact of social media
SOCIAL DATA BUSINESS DATA
On topic activity CRM behaviour
– Posts (volumes trends / time) – FAQ visits
– Multipliers (contagion/ formats) – Online questions
– Quality (likes/dislikes; sentiment) – Call centre ($$)
Sales behaviour
– search/research
– Micro site visits
– Promotion click thru
– Purchases ($$)
Investor behaviour
– Buy / Sell
–Associated impact ($$$)
ONE APPROACH TO BUSINESS IMPACT
Combines media data (unstructured) with
business data (structured) ; discovers
correlations between rising media topics and
changes in business data
HARVESTING QUANT PROCESSING QUAL FILTERS
Editorially NLP/Semantic engine, Verification &
Controlled modeling Analysis
SOCIAL DATA GUIDING PRINCIPALS
# 1. Use the broadest set of data for true perspective
# 2. Combine methodologies, then be consistent
# 3. Think scalable processes, not just big data
# 4. Align to business objective, deliver KPIs for
key stakeholders
# 5. Know the difference between indicative and
qualified outputs
28
GROUP CHALLENGE OBJECTIVE
• Your team is leading the social strategy of an important
new car launch across five European markets.
– Launched date: 1st. April 2013
– Social seeding campaign starting January 2013 led by
celebrity Cristiano Ronaldo
– Full national and network campaign from launch
• A key task is to create a set of Key Performance
Indicators to measure the campaign success in earned
social media
• To help with budget and buy-in, you
are meeting important internal
stakeholders. Tomorrow you will be
presenting the campaign KPIs to the
Marketing function
31
GROUP CHALLENGE OBJECTIVE
• Your team is leading the social strategy of an important
new car launch across five European markets.
– Launched date: 1st. April 2013
– Social seeding campaign starting January 2013 led by
celebrity Cristiano Ronaldo
– Full national and network campaign from launch
• A key task is to create a set of Key Performance
Indicators to measure the campaign success in earned
social media
• To help with budget and buy-in, you
are meeting important internal
stakeholders. Tomorrow you will be
presenting the campaign KPIs to the
Marketing function
32
GROUP CHALLENGE OBJECTIVE
• Your team is leading the social strategy of an important
new car launch across five European markets.
– Launched date: 1st. April 2013
– Social seeding campaign starting January 2013 led by
celebrity Cristiano Ronaldo
– Full national and network campaign from launch
• A key task is to create a set of Key Performance
Indicators to measure the campaign success in earned
social media
• To help with budget and buy-in, you
are meeting important internal
stakeholders. Tomorrow you will be
presenting the campaign KPIs to the
Marketing function
33
SOCIAL NETWORK METRICS (sample)
1. Number of followers
2. Number of following
3. Overall number of tweets
4. Average tweeting frequency (tweets/day based on last week/month)
5. Share of on-topic tweets (ratio of tweets on monitored topic vs. overall number of tweets for the study period – if applicable)
6. Average number of retweets (measures how engaged this users’ followers are)
7. Number of retweets for most retweeted posting (measures whether one hit-tweet was an outlier)
8. Share of posts with link to content (measures how much the user tends to tweet interesting content, vs. creating own tweets)
9. Share of tweets with @reference (measures how much the user is prone to conversations/interactions)
Facebook Groups:
Group level (Manual / not auto)
1. Closed vs. Open
2. Number of members
3. Wall activity: number of wall posts for specific time period; frequency of wall posts – open groups only
4. Wall activity: is it only group owner posting, or are members posting as well – open groups only
Single post level (all metrics below apply to open groups only):
1. Number of likes
2. Number of shares
3. Number of comments
4. Number of different users commenting
5. Number of comments per commentator
6. Additional layer of information - comments can be liked too, so where applicable, most liked comments/commentators
per group could be identified (currently purely manual unfortunately)
Facebook / Google + Personal Profiles: (Public profiles only)
Personal user profiles level
1. Number of friends
2. Number of subscribers
3. Active since
4. Average number of status updates per day/week
5. Average number of comments per status updates
6. Number of different users commenting on past X number of status updates (measures if the user interacts only with a small
group of more loyal friends, or many people engage with him/her)
7. Average number of other shares (links, photos, apps, locations etc) per day/week
8. Number of pages liked
9. Number of groups joined
10. Other applicable data – if necessary (number of photos uploaded, which apps the user is using)
34
SOCIAL NETWORK METRICS (sample)
Single status update/content share level:
1. Number of likes
2. Number of shares
3. Number of comments
4. Number of different users commenting
5. Number of comments per commentator
6. Is the post sponsored– now that you can pay to promote your own posts this is a metric we should bear in mind,
because it may explain higher values in the above quant metrics (conditional upon weather Facebook identifies
posts as sponsored!)
7. Number of likes by the user to the comments received by friends (to explain whose comments is the user
preferring/supporting)
LinkedIn
1. Number of Connections
2. Number of Companies in work experience
3. Number of Universities/Academic Institutions in education/training
4. Number of groups/professional communities joined
5. Number of status updates/shares
YouTube
Channel level:
1. Number of subscribers
2. Number of videos uploaded
3. Average frequency of uploading (videos per week/month
4. Average number of views per video
5. Average number of comments per video
Video level:
1. Number of views
2. Number of likes
3. Number of dislikes
4. Number of comments
5. Number of video responses
35
GROUP CHALLENGE OBJECTIVE
• Your team is leading the social strategy of an important
new car launch across five European markets.
– Launched date: 1st. April 2013
– Social seeding campaign starting January 2013 led by
celebrity Cristiano Ronaldo
– Full national and network campaign from launch
• A key task is to create a set of Key Performance
Indicators to measure the campaign success in earned
social media
• To help with budget and buy-in, you
are meeting important internal
stakeholders. Tomorrow you will be
presenting the campaign KPIs to the
Marketing function
37
Hinweis der Redaktion
Collective fear in businesses that they have not got a handle on social Multiple formats Too dynamic to measure Too voluminous to find relevancy RESULT: REPUTATION EXPOSURE / BUSINESS RISK
Econsultancy and AdobeIt is based on a survey of 650 marketing professionals.Technical issues:social data is stored in disparate tools (31%) social analytics are separate from multichannel analytics and business intelligence (31%)However agencies: Were more likely to point to a lack of budget or buy-in from the top of the organisation (40%) Lack of joined-up thinking (40%).Collective fear in businesses that they have not got a handle on social Multiple formats Too dynamic to measure Too voluminous to find relevancy RESULT: REPUTATION EXPOSURE / BUSINESS RISK
Fragmentation of data capture; AvinashKaushik Digital Evangelist for Google, Author of Web Analytics 2.0http://www.kaushik.net/avinash/best-social-media-metrics-conversation-amplification-applause-economic-value/Different apps give rise to different metrics; making holistic view challengingNew behaviours such as ‘pins’ in Pintrest“I feel we’re all data chemists at this point, trying to put a bunch of stuff into our beakers to see if it works” Ken Burbary, Chief Digital Officer at Campbell EwaldOrganizations move slowly: 63% of Alimteter survey reported 1-2 people tasked with social media measurement
Primary drivers are ‘customer insight’ – 84%However further inspection shows financial drivers make up most of the others – showing real focus in on ROIAli Ardalan, Intel‘why do you do an ROI analysis? To justify why you should do this project vs another, Why you need more funding. You need to know the results, are you wasting money?’ Could you have done the same thing with 20% less money?’Big Issue:Underlying expectation that AvinashKaushik talks about; Social media data should be a trackable as Web Analytics that have been so effective at optimizing online experiences, driver new dynamic models in advertising. But are they?The underlying tension here is the assumption that we can understand what people are thinking, wanting by the structured data elements in social data – such as ‘status’ updates, ‘likes/dislikes’, retweets and followers on twitter.As we shall see – these are ‘indicators’ and very valuable ones; however understanding what people are ‘saying’ and therefore thinking is more difficult and requires technologies that relay on processing ‘unstructured’ data – not ‘structured’
Mckinsey study:1,469 CEOs / C-suite April 2012 65% say dig data / analytics a top or top three corporate priority 68% say same about digital marketing & social tools 50% say current spending too small to deliver on aspirations of business transformationBudgets are expanding: 25% responded saying they spend at least 3% of total cost base on digital initativesBut lack of agreement between C-Suite:40% CEO expect investment at least 3%20% of CFOs agreed at this level, most say less!12% CIOs agree on this level, most say more!
Why is social media data difficult?In our dealings with clients, and I am sure this will be backed up by Maz later; there is currently what I call a new pain point paradigm – cheesy – but you are at the vortex of many elements that current are not ‘connecting’ – making your job very painful.Here are some of them – which I am sure you will be familiarBasically – there is no single remedy.Yet.But massive investments by Oracle, Salesforce and IBM point the way a tech architecture that will break down many of these data issues, and provide a single solution – or part…we are not there yet.
Best practice for EARNED media is to combine connect KPIs for Editorially Controlled & Socially Generatede.g. Opinions being gathered/refined/re-served in Editorial media vs. spontaneous or solicited opinions in social.KPIs need to connect to understand the relationship between the two formats
How does Klout work:has raised about US$40-million in funding to date.Klout now processes more than 2.7 billion pieces of information a dayalgorithm takes into account 400 distinct signalsGoogle, by comparison, only uses 200From from Twitter, now comes from Facebook, Google+, LinkedIn, Foursquare and Wikipedia It promotes three specific measures, which the company calls “true reach,” “amplification,” and “network impactDoes it capture the distinctions of actual, real world influence?How we think it works:The content you link to, and where you link to it fromThe conversations you have with people and their topicHow your followers act on your content (primarily how they retweet/share your content)Technology:mining/processing structured data: Process billions of ‘status’ updates on Facebook, Linked In, Twitter– est. 6 billion per day Process billions changes to ‘connections’ / followers – e.g. Social Graph – est. 4 billion per day Processing unstructured data: From 100+ million pieces of unstructured content – such as tweets, FB comments, linked inWe derive hundreds of thousands of different topics that 14 million users are influentialonOn average 5 topics per user using NLP and semantic analysis NLP (statistical) tags 1 million topics / ontology Key thing:Combine changes in ‘status’ and ‘connections’ to ‘topics’ (as they are defined)3 months of mentions and retweets are analyzed, currently over 6 billionRecent comments from Joe Fernandez:We are constantly evolving the algorithm with an eye toward increasing the score’s accuracyWe define influence as the ability to drive action. A person’s ability to get their network to engage with them is much more telling than their follower count.Through Klout, brands provide special offers, called “Perks,” to influencers and pay a fee to make those offers. We are also considering other forms of revenue, but Perks makes up the lion’s share of our income.
How does Klout work:has raised about US$40-million in funding to date.Klout now processes more than 2.7 billion pieces of information a dayalgorithm takes into account 400 distinct signalsGoogle, by comparison, only uses 200From from Twitter, now comes from Facebook, Google+, LinkedIn, Foursquare and Wikipedia It promotes three specific measures, which the company calls “true reach,” “amplification,” and “network impactDoes it capture the distinctions of actual, real world influence?How we think it works:The content you link to, and where you link to it fromThe conversations you have with people and their topicHow your followers act on your content (primarily how they retweet/share your content)Technology:mining/processing structured data: Process billions of ‘status’ updates on Facebook, Linked In, Twitter– est. 6 billion per day Process billions changes to ‘connections’ / followers – e.g. Social Graph – est. 4 billion per day Processing unstructured data: From 100+ million pieces of unstructured content – such as tweets, FB comments, linked inWe derive hundreds of thousands of different topics that 14 million users are influentialonOn average 5 topics per user using NLP and semantic analysis NLP (statistical) tags 1 million topics / ontology Key thing:Combine changes in ‘status’ and ‘connections’ to ‘topics’ (as they are defined)3 months of mentions and retweets are analyzed, currently over 6 billionRecent comments from Joe Fernandez:We are constantly evolving the algorithm with an eye toward increasing the score’s accuracyWe define influence as the ability to drive action. A person’s ability to get their network to engage with them is much more telling than their follower count.Through Klout, brands provide special offers, called “Perks,” to influencers and pay a fee to make those offers. We are also considering other forms of revenue, but Perks makes up the lion’s share of our income.
How does Klout work:has raised about US$40-million in funding to date.Klout now processes more than 2.7 billion pieces of information a dayalgorithm takes into account 400 distinct signalsGoogle, by comparison, only uses 200From from Twitter, now comes from Facebook, Google+, LinkedIn, Foursquare and Wikipedia It promotes three specific measures, which the company calls “true reach,” “amplification,” and “network impactDoes it capture the distinctions of actual, real world influence?How we think it works:The content you link to, and where you link to it fromThe conversations you have with people and their topicHow your followers act on your content (primarily how they retweet/share your content)Technology:mining/processing structured data: Process billions of ‘status’ updates on Facebook, Linked In, Twitter– est. 6 billion per day Process billions changes to ‘connections’ / followers – e.g. Social Graph – est. 4 billion per day Processing unstructured data: From 100+ million pieces of unstructured content – such as tweets, FB comments, linked inWe derive hundreds of thousands of different topics that 14 million users are influentialonOn average 5 topics per user using NLP and semantic analysis NLP (statistical) tags 1 million topics / ontology Key thing:Combine changes in ‘status’ and ‘connections’ to ‘topics’ (as they are defined)3 months of mentions and retweets are analyzed, currently over 6 billionRecent comments from Joe Fernandez:We are constantly evolving the algorithm with an eye toward increasing the score’s accuracyWe define influence as the ability to drive action. A person’s ability to get their network to engage with them is much more telling than their follower count.Through Klout, brands provide special offers, called “Perks,” to influencers and pay a fee to make those offers. We are also considering other forms of revenue, but Perks makes up the lion’s share of our income.