Big Data and Social Analytics - at IBM's Information on Demand Conference. Aya Soffer | Director, Information Management & Analytics Research & Mark Heid | Program Director, Social
Analytics
BSC 3362 - Big Data and Social Analytics - IOD Conference (IBM)
1. Revolutionizing How Business
Understands Customers -- Big
Data Meets Social Analytics
Session Number BSC-3362
Aya Soffer | Director, Information
Management & Analytics Research | IBM
Mark Heid | Program Director, Social
Analytics | IBM
#ibmiod #ibmiod
2. Please note
IBM’s statements regarding its plans, directions, and intent are subject to change or
withdrawal without notice at IBM’s sole discretion.
Information regarding potential future products is intended to outline our general
product direction and it should not be relied on in making a purchasing decision.
The information mentioned regarding potential future products is not a commitment,
promise, or legal obligation to deliver any material, code or functionality. Information
about potential future products may not be incorporated into any contract. The
development, release, and timing of any future features or functionality described
for our products remains at our sole discretion.
Performance is based on measurements and projections using standard IBM benchmarks in
a controlled environment. The actual throughput or performance that any user will
experience will vary depending upon many factors, including considerations such as the
amount of multiprogramming in the user’s job stream, the I/O configuration, the storage
configuration, and the workload processed. Therefore, no assurance can be given that an
individual user will achieve results similar to those stated here.
#ibmiod
3. Agenda
1 Our Perspective on Big Data Analytics
2 A Look at Big Data Social Analytics
• Multi-channel Marketing
• Customer Care and Insight
• End-to-End Demo
3 IBM Research: Driving the Revolution in Big
Data Social Analytics
3
#ibmiod
4. We’ve Moved into a New Era of Computing
12 terabytes 5 million
of Tweets trade events
create daily per second “We have for the first time
an economy based on a
key resource
Volume Velocity
[Information] that is not
only renewable, but self-
generating.
Variety
Running out of it is not a
Veracity
100’s problem, but drowning in
it is.”
Of video feeds from
surveillance cameras
– John Naisbitt
4
#ibmiod
5. Challenges of Big Data – The New Mix of Information
Enterprise Data Machine Data Social Data
• Volume • Velocity • Variability
• Structured • Semi-structured • Highly unstructured
• Throughput • Ingestion • Veracity
5
#ibmiod
6. Typical Client Use Cases with New Types of Analytics
Compute
Intensive Gain more complete
• Fraud Detection answers to business
• Smart Grids and Smarter Utilities decisions to make
better decisions faster
• Risk Management and Modeling
Ask new questions
• Asset Management and Optimization
about their business to
• Call Detail Records uncover new value or
• Call Center Transcripts realize cost-savings
• Log Analytics
Explore and
• 360°View of the Customer experiment to find
• Data Warehouse Evolution new opportunities and
Storage create new business
Intensive models
6
#ibmiod
7. IBM Big Data – Analytics and Platform
IBM Big Data –
Analytics and Platform
• Addresses 4Vs of information
Visualize and Experiment
Predict Analyze Real-time
• Harnesses the next wave of
analytics that exploits value
from a rich information mix
Search and Discover
Hadoop Stream Data
• Fosters a new era in analytical System Computing Warehouse
applications
Integrate and Govern
7
#ibmiod
8. Most Client Use Cases Combine Multiple Technologies
Pre-processing
• Ingest and analyze unstructured data types
and convert to structured data
IBM Big Data - Combine structured and unstructured analysis
Analytics and Platform
Visualize and Experiment
• Augment data warehouse with additional external
Predict Analyze Real-time
sources, such as social media
Search and Discover
Hadoop
System
Stream
Computing
Data
Warehouse Combine high velocity and historical analysis
• Analyze and react to data in motion; adjust models
Integrate and Govern with deep historical analysis
Reuse structured data for exploratory analysis
• Experimentation and ad-hoc analysis with structured
data
8
#ibmiod
10. Agenda
1 Our Perspective on Big Data Analytics
2 A Look at Big Data Social Analytics
• Multi-channel Marketing
• Customer Care and Insight
• End-to-End Demo
3 IBM Research: Driving the Revolution in Big
Data Social Analytics
10
#ibmiod
11. Even though social media is pervasive, using it successfully in
marketing campaigns today is hit or miss
Measurement and ROI are
elusive
Campaigns are poorly About half of marketers
integrated admit that their social
Only brand / mass marketing media marketing efforts
techniques are employed
Opportunity to engage
are totally siloed
individuals is ignored
Source: Q4 2010, Unica’s Global Survey of Marketers
1111
#ibmiod
12. By linking together social and customer data, we can help our clients
market more effectively across multiple channels
Planning, coordinating and executing marketing campaigns
to stimulate demand – it’s a process that includes social media
Insights from Create Optimize email, display Deliver targeted
social media relevant and search ad programs messages and offers
and other messages
data sources
Capture & analyze
responses and
refine
1212
#ibmiod
13. Introducing: Multi-channel campaign management with integrated
social analytics
An integrated approach which allows organizations to measure, adjust and, ultimately,
use social media data to gain greater precision for their campaigns.
How can I leverage • Measure the social impact
social analytics to optimize of campaigns through
return on my campaigns? earned and owned media
Ma rke ting • Gain greater campaign
Ma na ge r
precision by applying
predictive models to
socially-derived segments
How can I maximize the • Evolve and align
value of our social insights marketing and social
for marketing? campaigns through a
S oc ia l Me dia centralized workspace
Ana lys t
13
13
#ibmiod
14. Big Data Social Analytics in
Social Business & Smarter
Commerce
14
#ibmiod
15. “Benjamin's Grocery” - Winning with Social Analytics & Smarter Commerce
How does it work?
Analytics Emerging Topics Affinities
Conversations you asked What is correlated with what?
Sentiment dashboard
about and those you didn't
Perceptual Map
Social Media Spatial alignment of attributes
• Tweets
• Blogs
• Forums
Communities
1 Derive ideas, insights and
• Surveys
• Advocate dialog
• Discussions
actions from Social Media
2 Pulling consumers from where the conversation is
on the web, match them to segments based on
their actions on Benjamin's website
Customer
3 Execute the campaign using Individual
Data for consumers who opted-in
Website
Behavior
• Clicks
• Searches
Previous
• Views
Campaign Data
• Contact history
• Response/purchases
• Test campaigns
Modeling Scoring Campaigns
Predict who is likely to Rank best offers Multi-Channel Marketing
15 respond
#ibmiod
16. “Benjamin's Grocery” - Winning with Social Analytics & Smarter Commerce
What is the storyline?
Introducing Benjamins Grocery Stores Competition in the grocery business
can be intense and Benjamins faces their fair share with Jurassic, a low-price chain with
broad presence in the market.
The Market Event On January 20th, 2012, Jurassic announces the end of ad hoc
campaigns and the beginning of “every-day low prices”. They drop prices by 12-15% for
3000 products.
Benjamins' Research Knowing that they can't profitably copy Jurassic's price
strategy, Benjamins mobilizes a team of experts to search for a better response. They
discover that customers have a core un-met need for “healthy, interesting meals at a
fair price”.
Benjamins' Response The Benjamins team rapidly tests a creative plan to hire
well-known chefs to sponsor new recipes that use Benjamins store brand products. Their
communities-of-interest like it – particularly “Moms”, “Singles” and “Gourmets”. They
kick-off a new 1:1 cross-channel campaign that lasts through the rest of Q1.
The Results Over the two-month campaign, Benjamins gains market share and grows
profit by 8%.
16
#ibmiod
17. “Benjamin's Grocery” - Winning with Social Analytics & Smarter Commerce
What products are used?
Analytics Emerging Affinities
Where can all ofSentiment dashboard
the Conversations you asked
Topics What is correlated with
How can Benjamin's quickly
about and those you didn't
what?
relevant information be understand their differentiatorsPerceptual Map
and
Social Media brought together for competitor vulnerabilities? Spatial alignment of
• Tweets
• Blogs
productive decision- attributes
• Forums making? What can they use to do root cause
Communities
analysis and uncover un-met needs
1 Derive ideas, insights
• Surveys
• Advocate dialog
among their target customers?
• Discussions
and actions from Social
Media
2 Pulling can Benjamin's pivot from conversation is
How consumers from where the
aggregate to individual data?
on the web, match them to segments based on
their actions on Benjamin's website
3
What optimization can beusing
Execute the campaign applied
Customer to campaign parameters?
Individual Data for consumers who
Website
Behavior opted-in
• Clicks
• Searches
Previous
• Views
Campaign Data
• Contact history
• Response/purchases
• Test campaigns
Modeling Scoring Campaigns
Predict who is likely to Rank best offers Multi-Channel Marketing
17 respond
#ibmiod
18. “Benjamin's Grocery” - Winning with Social Analytics & Smarter Commerce
What products are used?
Analytics Emerging Affinities
Conversations you asked
Topics What is correlated with
Sentiment dashboard
about and those you didn't what?
Perceptual Map
Social Media Spatial alignment of
• Tweets attributes
• Blogs
• Forums
Communities Cognos Consumer Insight 1.1
●
1 Derive ideas, insights
● SPSS Modeler 15.0
• Surveys
• Advocate dialog
• Discussions
and actions 10.1 Social
● Cognos from
Media
● Connections 4.0
2 Pulling consumers fromAnalytics conversation is
● Coremetrics Web where the
● on the web, match them to segments based on
Cognos Consumer Insight 1.1
their actions on Benjamin's website
● Unica Campaign
Customer
3 Execute the campaign using
● SPSS Modeler 15.0
Individual Data for consumers who
● Cognos Consumer Insight
Website
Behavior opted-in
• Clicks
• Searches
Previous
• Views
Campaign Data
• Contact history
• Response/purchases
• Test campaigns
Modeling Scoring Campaigns
Predict who is likely to Rank best offers Multi-Channel Marketing
18 respond
#ibmiod
20. Business Analytics and Big Data Platform Integration
Business Analytics
SPSS Cognos Cognos Cognos CCI
Predictive RTM BI Insight
Predictive Real-time Reporting / Analysis Export and Unstructured
Analytics Dashboards Explore Analysis
InfoSphere
BigInsights
InfoSphere Data
Streams Warehouse BigSheets BigIndex Hive HBase
Hadoop (Map-reduce)
File system (GPFS, HDFS)
Load through UDFs
20 IBM Confidential: References to potential future products are subject to the Important Disclaimer provided earlier in the presentation
#ibmiod
21. Agenda
1 Our Perspective on Big Data Analytics
2 A Look at Big Data Social Analytics
• Multi-channel Marketing
• Customer Care and Insight
• End-to-End Demo
3 IBM Research: Driving the Revolution in Big
Data Social Analytics
21
#ibmiod
22. Social Analytics in IBM Research - moving up the value stack to
extract actionable insight
Filtering social media is Summarization is critical in
challenging and critical Relevance Filtering Topic Modeling diffuse content streams)
Information Summarization
Needs to be multi-lingual Detecting intent to buy or intent to
and tuned to specific Sentiment Lexical Pattern Extraction act or mood or brand attributes
domains
Lexical Extraction
Discover hidden pockets of
Influence is critical component for Influence Community Detection expertise in an enterprise setting
social media filtering and
Enterprise expertise
Influence and Communities
Extract customer demographic Context (eg location) is key
features that can be joined with Customer Modeling Situational Context
differentiator in an increasing
legacy attributes number of applications
22 User Modeling
#ibmiod
23. Social Pulse
Social Pulse – What are employees saying about their
company’s brand
• A Social Analytics Solution for marketing and communications
professionals
• Focuses on internal versus external consumer perception of
your brands and products
• Based on the idea of your workforce being brand
ambassadors
• Experimenting within IBM
• Externally
>25,000 employees on Twitter, >300,000 on LinkedIn, and > 198,000 on
Facebook
• And Internally
> 300,000 IBMers use IBM Connections Communities, Blogs, Wikis,
Profiles, Forums etc.
23
#ibmiod
24. The Users Social Pulse
What brand
related topics are
IBMers talking
about this week? everyone on
Is
board with our new
Smarter Planet
strategy?
Which business
units get the
message, which
ones are still
struggling?
Are our
management teams
helping our brands
to be presented in
the best light?
24
#ibmiod
24
26. By Business Unit & Common Topics
Across Business Units
Search for brand
specific topics
26
27. Not All Business Units are Positive
Let’s see if there are
differences across countries
Within S&D
27
28. S&D Ireland Very Positive, Opening New
Technology Center, Ireland Research (= new
Technology Center) is reserved.
28
29. Brandy
Brandy – Associating brand perceptions with customer traits
Mining of customer traits
• Demographics
[Ford, 2005]
• Personality
• Fundamental needs
• Preferences
•…
• Integrating mined inv
s. co ent
information with existing u sv ns ive
vo ist /c
er ent en u ri
customer data e/n fid t/c ou
au s v
itiv /con t io s .
ns
se cure us
se
• Associating brand
frie s. col
d
ate
nize
perceptions with customer
ndly
v
vs. e nt/orga
ss
/com /unkin
traits especially their
rele
asy-
d
pas d
g/ ca
“needs map”
ie
effic
sion
goin
outgoing/energetic vs.
29
solitary/reserved #ibmiod
30. Brandy
Example: Modeling and Deriving Personality
Map the use of words, frequency, &
correlation with Big5 based on LIWC
“Agreeableness”
wonderful (0.28), together (0.26) …
porn (-0.25), cost (-0.23)
Openness
Conscientiousness
Extraversion
Agreeableness
Neuroticism
0% 20% 40% 60% 80%
[Tausczik&Pennebaker 2010, Yarkoni
30 2010]
#ibmiod
32. Campaign management: a Retail Example Brandy
Help Retailer identify customer segments to launch “
CoolBrand” collection
Openness: 83% Openness: 23%
Idealist: 62% Realist: 87%
Interest: Dining Interest: Travel
50% close ties: openness 75% 35% close ties: interested in travel
… experience fine dining at … Want your luggage to stand out
home in Italian fashion style: at the airport? Never need to dust
“CoolBrand” dinnerware… it? Here comes “CoolBrand”
collection…
Save 5% by sharing this with
your 5 (open-minded) friends Save 5% by sharing this with your 5
such as … (travel-loving) friends such as…
32
#ibmiod
33. A Smarter Cities Example Brandy
Help DMV identify suitable segments for
different campaigns
Conscientiousness: 23% Neuroticism: 53%
Realist: 92% Idealist: 71%
Interest: Foodies Interest: Travel
50% close ties: Conscientiousness 25% 35% close ties: interested in travel
… Holiday is around the corner … Your current insurance policy
… is up for renewal …
Here are holiday safe driving tips:
http://dmv.ca.gov/... Share this with your 5 (travel-
loving) friends such as… and ask
share this with your close friends them to follow us to receive
33
such as … reminders…
#ibmiod
34. COPS
COPS – Crowdsource Oriented Public Safety
Automatic detection of Public Safety incidents and KPIs, from
crowdsourcing data, which is incomplete, inaccurate and noisy
Emergencies, Limited
call for help coverage
Use innovative “fusion analytics” to reliably detect incidents and
trends from uncertain data, textual, spoken and numerical
Analytics
• Event / fact
Crowd and fusion
source summarizations
(voice
in near- • KPIs
& text) real-time
Social
media sensors
34
#ibmiod
35. COPS
Sample Use Case (Managing Natural Disasters)
Event 1 – 10:10 river water surging
(from accumulation of tweets)
Event 2 – 11:15 fast moving
water (from accumulation of Event 3 – 11:15 – flood, major
mobile messages) road blocked (from accumulation
of tweets and mobile messages)
Event 4 – 12:30 – flood (from
Event 5 – 12:30 – traffic accumulation of tweets and
accident (from accumulation mobile messages)
of mobile messages)
35
#ibmiod
36. COPS
System automatically aggregates and filters the data
Crowd-source events that reflect aggregated data – to
avoid overloading Event 1 – 10:10 river of crowd-source data
by large volume water surging
and to reduce uncertainty by fusing tweets) posts
(from accumulation of multiple
Crowd-source events that are progressive – updated as
Event 2 – 11:15 fast crowd-source data becomes available
more moving
water (from accumulation of Event 3 – 11:15 – flood, major
mobile messages) road blocked (from accumulation
of tweets and mobile messages)
Crowd-source events that display the inherent uncertainty
(confidence) – from the event4description to(from location
Event – 12:30 – flood
the
Event 5 – 12:30 – traffic accumulation of tweets and
accident (from accumulation mobile messages)
of mobile messages)
36
36
#ibmiod
37. COPS
Main Module - Event Profile Generation
(1) Data Ingestion filter (4) Event Detection
relevant information from Statistical detection &
millions of messages model-based detection
Filters
Data Statistical (5)
patterns Reporting/Alerting/D
ingest
ashboarding
Fuse & Event Detection
Unstructured Aggregate
data sources
Streams / BigData Platform
Event Events, event
Entity/ representation summaries, trends,
Event
Extraction KPIs, Predictions
Join/Fuse
/Aggregate
BigInsights /BigData Platform Event Schema
(2) Extraction/Integration (3) Automatic Model
Flow from Generation from
unstructured data entity schema to
(tweets and crowd Event model on
data) to JSON objects BigInsights
3737
#ibmiod
38. Microcosm
Microcosm - uncover the commercial potential of local
microcosms
• Understand the marketing potential of particular locations beyond the
individual level
• Understand the potential of viral marketing
• Identify promising community types and target marketing to them
• Lower marketing costs by targeting earned media
38
#ibmiod
39. Microcosm
Social Analytics to extract communities and Locations
Extended community Identifying participants location
of people that talk about based on profiles and discussions
some subject
39
#ibmiod
40. Microcosm
Geographical Analytics – How it works
• GPS Geotagging (<5% of tweets)
• Even if explicit in profile – disambiguation might be needed:
• E.g., “Springfield” by itself can refer to 30 different cities in the USA.
• Techniques used
• Rule-based
E.g., “I live in ..”, “lets meet at ..”
• Machine learning (supervised):
Statistical methods- find the most characteristic terms of people
that report they live in some location x.
E.g., “The Strip”, “Bellagio fountains”, “Freemont St.”…-> Las
Vegas
• Based on Social Network,
• i.e. learn location of people
based on the locations of their friends
Location 1 Location 2 Location 3
40
#ibmiod
41. Microcosm
Community Analytics - How it works:
How we build the communities:
• Build social graph based on the data flow in the social media. For
example, in Twitter, using the @Reply tag.
• Extend the connections with friends, followers, following, etc.
• Then use clustering-based approach
What we gain from the communities analysis?
• which features have commercial significance
• which features can be acted upon
41
#ibmiod
42. Thank You!
Your Feedback is Important!
• Access SmartSite to complete your session surveys
o Any web or mobile browser at iodsmartsite.com
o Any SmartSite kiosk onsite
o Each completed session survey increases your chance to win
an Apple TV with daily drawing sponsored by Alliance Tech
#ibmiod