Social Media Intelligence

Wendy W. Moe wmoe@rhsmith.umd.edu

www.wendymoe.com @wendymoe

David A. Schweidel dschweidel@emory.edu

www.dschweidel.com @dschweidel

• The challenge facing organizations

• Social Media Intelligence – Understand the science of opinion and the implications for

social media data and metrics – Measure what matters (not what’s easy to measure)

What is Social Media Intelligence?

2

Building Social Media Intelligence

3

WHAT?

WHERE?

WHO?

“Online Product Opinions: Incidence, Evaluation and Evolution,” Marketing Science, 2012, with David A. Schweidel

“The Value of Social Dynamics in Online Product Ratings Forums,” Journal of Marketing Research, 2011, with Michael Trusov

“Sentiment and Venue Choice in Social Media Posting Behavior,” under review at JMR, with David A. Schweidel

“Drivers of Social Media Rebroadcasting: Investigating the Role of Message Content and Influencers,” under review at JMR, with Yuchi Zhang and David A. Schweidel

INTE

GRAT

ION

Soci

al M

edia

Inte

llige

nce

by

Wen

dy W

. Moe

and

Dav

id A

. Sch

wei

del,

Cam

brid

ge U

nive

rsity

Pre

ss, c

omin

g Fa

ll 20

13

4

WHAT do we say?

Social Media Posting Decisions

Opinion / Brand Evaluation

Do I post? What do I post?

Where do I post?

SentimentProduct

Attribute

Venue FormatDomain

5

Rating Trends

Variance

Average

Activists

Post frequently Attracted by lack of consensus More negative Variance and volume make them more negative

Low Involvements

Post infrequently Deterred by lack of consensus More positive Variance and volume make them more positive

6

Implications

7

What should we do?

• If you’re listening, don’t over-react. The ratings environment does not necessarily reflect the opinions of the entire customer base

• “Listen to ALL the American people” – Encourage diversity of opinion – Monitor different venues / data sources

8

9

WHERE do we post?

Social Media Posting Decisions

Opinion / Brand Evaluation

Do I post? What do I post?

Where do I post?

SentimentProduct

Attribute

Venue FormatDomain

10

What affects posted sentiment?

General Brand Impression (GBI)

Venue Venue-specific dynamics Message topic

11

Does it Matter Where We Listen?

0%

20%

40%

60%

80%

100%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Dist

ribut

ion

of C

omm

ents

Observation Month

Blog

Forum

Microblog

Other

0%

20%

40%

60%

80%

100%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Prop

ortio

n of

Pos

itive

Com

men

ts

Observation Month

Blog

Forum

Microblog

Aggregate

12

From a sample of social media comments pertaining to an enterprise software brand (courtesy of Converseon)

Plus, sentiment metrics vary depending on topic

13

GBI vs. Offline Survey • Potential for GBI as a lead

indicator

• Correlation with survey (t) – GBI = .376 – Avg sentiment =.008 – Blogs = .197 – Forums = -.231 – Microblogs = .394

• Correlation with survey (t-1)

– GBI = .881 – Avg sentiment = .169 – Blogs = .529 – Forums = .213 – Microblogs = .722

8.75

8.8

8.85

8.9

8.95

9

9.05

9.1

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

1 2 3 4 5 6 7 8 9 10

Aver

age

Surv

ey R

espo

nse

GBI

Month of Overlap Period (t)

GBI in month t-1 Survey in month t

14

GBI vs. Monthly Close of Stock Price

Mean Coeff % p-val <.05

Constant -68.09 0.16

S&P* 0.10 1

GBI(t) -16.43 0.01

GBI(t-1) 30.81 0.99

Adj R-sq .46

15

Regressed brand’s stock price against GBI (controlling for market index). Across iterations…

Social Media Intelligence vs. Social Media Noise

• Account for differences in social media sentiment across venues and across sub-topics

• Potential to use social media for market research – Adjusted measure (GBI) can serve as lead indicator

• At a minimum, dashboard metrics should disaggregate and

separately report venue-specific and topic-specific metrics

16

Questions?

Why Do People Post? Opinion Formation versus Opinion

Expression

• Opinion formation

– Pre vs. post purchase – Customer satisfaction and word-

of-mouth

• Opinion expression

– Opinion dynamics – Opinion polls and voter turnout

Pre-Purchase Evaluation

Purchase Decision and Product Experience

Post-Purchase Evaluation

Incidence Decision

Evaluation Decision

Posted Product Ratings

EXPE

RIEN

CE M

ODE

L IN

CIDE

NCE

& E

VALU

ATIO

N M

ODE

LS

SELECTION EFFECT

ADJUSTMENT EFFECT

18

Perils of Selective Listening

• Index of over/under-estimation of messages mentioning specific product attributes

Application Compatibility

Documentation/Support Price Ease of Installation

Brand Reputation

Blog only -27.73 20.06 18.72 -31.39 13.43

Forum only 81.71 -21.20 -6.32 74.03 -31.99

Microblog only -57.37 6.02 -7.50 -46.90 22.00

Blog+forum 34.25 -3.30 4.54 28.32 -12.29

Blog+microblog -44.08 12.32 4.26 -39.94 18.16

Forum+microblog 14.26 -8.00 -6.89 15.38 -5.81

Blog+forum+microblog 2.38 -.06 .35 2.15 -0.36

19