1. Data-Mining Social
Audiences to Impact a
Brand’s Marketing Mix
Matt Esslinger, COO of XA.net (matt@xa.net)
October 21, 2011
Confidential
Confidential
2. Social Media and Data: Biggest Problems for CMOs
Brands see shifting user demographics and need to use social data - and take user feedback into account earlier -
in the market planning and product processes
Source: IBM Global Chief Marketing Officer Study (n=1,734)
Confidential
Confidential
3. Social Media + Analytics: CMOs to Increase Spend!
Chief Marketing Officers of the world’s largest companies globally have been taken by surprise by the explosion
in consumer data and the transformative potential of social media – and need technology to address it
% of CMOs reporting under preparedness Plans to increase the use of technology
Source: IBM Global Chief Marketing Officer Study (n=1,734), Oct 2011
Confidential
Confidential
4. Facebook Targeting: The Big Picture
US Facebook audience
addressable by:
Gender 99%
Age 95%
Status 63%
Education 36%
Workplace 6%
Confidential
Confidential
5. Facebook Targeting: Some Highlights
• Younger people are more addressable than older people:
• A 14 year-old is twice as likely as a 45 year-old to like something on Facebook
• Women are more addressable than men:
• 19% more likely to share interests; 13% more likely to be married; 6% more likely to be engaged
• Exception: men are 12% more likely to list as single
• Addressable data is sticky, and sticky across types:
• Users who like any single keyword overindex for 94% of all other keywords
• Users who list a workplace overindex for 92 % of all keywords
• Users who list a workplace are 24% more likely to share a relationship status
Confidential
Confidential
6. Likes & Interests: Age Distribution
Young people are much easier to address with keywords than older people
Interest Addressability Index by Age
200
150
100
50
-
13 18 23 28 33 38 43 48 53 58 63
Age
Confidential
Confidential
7. Likes & Interests: Reach Distribution
Few keywords have very large audiences, and many keywords have small audiences
Keyword Count by Audience Size (log-log)
100,000
10,000
Mid-Tail
1,000
100
Popular
10
1
1,000 10,000 100,000 1,000,000 10,000,000
Audience Size
Confidential
Confidential
8. Data = The Currency of Advertising and Attention
We are the first company to bring together rich data from both social media (ads, apps etc.) + real-time display sources
and give marketers a way to take immediate action to reach those users with the optimal message
Advertiser targeting 3rd party Interest/demographics
25-34
Female
Proprietary Audience
Response Database
Portland
25-34 Female
Portland Social Predict TM
Strollers
Hyperlocal data
Ad targets + category intent
Confidential
Confidential
9. SocialPredict Provides a Unique Data Foundation
XA.net uses 3+ years of historical social graph data to make audience predictions
Content-Specific Metadata
…creates structure among…
650,000+ Targetable Interests
…of users organized into…
300,000 Networks
…providing a view into…
SocialPredict
…which has 370 million unique
edges and is growing every day…
Confidential
10. Optim.al: Proprietary Audience Segmentation
Gardening Gardening
Automated Suggestions garden olive garden
Data from millions of
plants olive garden bread sticks
optim.al vs. Facebook
landscaping plants
antiques bread
sewing
home improvement
bird watching
gardener audience profiles
Swimming Swimming
Proprietary Facebook
lifeguards swimming pool
water polo olympic swimming
tanning lifeguarding
diving
tubing
hot tubs
hot pools
water polo
triathalon
datasets combined with
More. Chess
scrabble
Chess
nil
external data:
Relevant. backgammon
monopoly
Targeting. Company Lists
ping pong
mathematics
card games
Musician Musician Likes & Interests
recording Anthony Hamilton
performing
guitarist
Saul Williams
Neo-Soul
Schools & Majors
drummer
producer Semantic Scrapes
songwriter
Ayn Rand Ayn Rand
Atlas Shrugged Atlas Shrugged
The Fountainhead
Anthem
Libertarian
Objectivism
George Orwell
Confidential
Confidential
11. Facebook: Keyword Selection Process
1. Translate an AdWords list from purchase intent to people
a. NLP: tokenization and stemming to facilitate Facebook keyword discovery
“realtor” “realtor marketer”, “national association realtors”…
b. Categorical Expansion: amplify with vertical keywords
“Bloomberg” “CNBC”, “Motley Fool”, “Yahoo Finance”…
c. Audience Suggestion: add highest indexing keywords among a core user base
“Barack Obama” “The Daily Show”, “NPR”, “Coffee”
2. Filter the keyword list, based on stated reach and CPC objectives
3. Slice into minimally overlapping segments and create a controlled environment
for testing and optimization
Confidential
Confidential
12. Facebook: Keyword Selection Process – Proof
Targeting Method CTR Conversion Rate CPA
Facebook Broad Category:
Small Business Owners .015% 3.9% $28.65
Search Keyword-Driven
(Intent+Behavior) Precise
Interests
.033% 11.6% $10.71
1. Broad Category Targeting (BCT) does not pass the smell test
• BCT identifies165,000 Small Business Owners under the age of 17
2. Only Precise Interest Targeting allows for deep relevance and segmentation
3. Search keywords (properly translated) can identify highly-qualified,
precisely target-able audience segments
Confidential
Confidential
13. Our Proprietary Data: A Visual Social Graph
http://optim.al/blog/
Confidential
Confidential
14. What Does All This Mean to a Brand?
Data drives their Strategy and Execution, both Online and Offline
A luxury auto brand learned that most
category fans are young + aspirational & Online
not likely buyers, but they also found the Strategy
loyal owner segment within their base
Analyzed millions of US fans to discover
interest-based clusters of underserved Offline
consumers by ZIP Code. These deep and Strategy
surprising insights were revolutionary
Confidential
Confidential
15. Data-Mining Social
Audiences to Impact a
Brand’s Marketing Mix
Matt Esslinger, COO of XA.net (matt@xa.net)
October 21, 2011
Confidential
Confidential