Professional Documents
Culture Documents
Media activities generated by consumers or communities that are neither paid for
Insead
nor induced by brand owners are claimed to have a potentially game-changing impact
marcel.corstjens@
insead.edu
Andris Umblijs
on communication and brand building. In this study, the authors propose a rigorous
methodology to assess the impact of this type of social media activities on the actual
McKinsey
andris_umblijs@
process to condense the complex reality of micro-social-media events for a brand into a
mckinsey.com
manageable set of social media indicators (SMI). These SMI subsequently are used as a
subset of the drivers, together with more traditional marketing-mix elementsin a general
market-response modelto estimate their relative impact on brand performance in the
market. This methodology is illustrated with two real-world examplesone in the market
for flat-screen-television market and the other in the set of Internet broadband-service
providers.
INTRODUCTION
DOI: 10.2501/JAR-52-4-433-449
brands.
media activities:
established
analyze
traditional
well-
marketing-mix
request; and
profitability.
mix elements.
media.
added:
434 JOURNAL
television measure.
determination.
contribution:
focus on
activities, and
company.
SM Listening
Results
Weighting/aggregation
methodology
Weekly SMR
(SM Negative)
Sales
Marketing
Stimuli
SM Positive
Promotional Price
SM Negative
POS Advertising
TV Advertising
Competitive
advertising
Base
Time (3 years)
Stage I: Define
436 JOURNAL
so on;
Technology: Plasma, LCD, LED, 3D,
and so on;
Define
Listen
Weight
Aggregate
Followers
KOL
(+)Sentiment
Topic
relevance
SubCategory 1
Topic 1
(1)Sentiment
Keywords/
Phrase
Topic
()Sentiment
Keywords/
Phrase
Topic n
Topics
unique
followers
Keywords/
Phrase
Category
SubCategory 2
SubCategory n
Keywords/
Phrase
Keywords/
Phrase
Figure 2 Methodology for the Conversion of Large Numbers of Individual Social-Media Events into
Social-Media Rating (SMR) Indicators
Brand Image: Brand A, Brand B (com-
neutral;
particularly relevant:
author
regarding
in
the
status
of
social-media
communitykey
opinion
leaders
communication.
creates; and
tion topic.
media
munication originator;
listening-service
providers
Table 1
Illustrative Example of the Weighting Procedure (Messages in the Table are for Demonstration
Purposes)
Stage of weighting
A
KOL
weight
Chain
Reach reaction
Topic
relevance
Total
score
Message
Category
Message
Sentiment count
Technology
Positive
36
Positive
27
Technology
Negative
Technology
Negative
Price
Negative
Price
Positive
18
Price
Negative
81
Positive
12
438 JOURNAL
apply
additional
weight
depend-
communication;
on social-media networks.
this
methodology
involved
history.
Subsequently,
these
variables
are
sales.
channels.
companies.
TELEVISIONS
televisionswith
prominent
market
(Table 2).
agency);
agency);
television,
Internet,
radio, and
outdoor.
agency); and
print,
Table 2
Descriptive Statistics of Model Variables
Mean Units
Minimum
Units
Maximum
Units
Standard
Deviation
9,316
3,672
45,832
6,937
Sales Units
Retail
486,748
31,235
1,671,086
369,207
Internet
445,222
3,374
898,531
292,601
157,902
1,397,164
350,366
TV
411,208
2,933,487
662,219
Radio
41,941
455,241
115,273
Outdoor
12,724
450,266
66,113
Disposable Income
8,822
8,666
9,135
117
45
26
61
11
Price
739
600
854
52
301
1,800
401
91
984
146
68
589
316
440 JOURNAL
CCI
where
and price.
marketing j j = ii (marketing j )t i
media.
2011):
Salest = 0 t + i market i
+ j (marketing j j ) j
(1)
Table 3
Model Independent Variable Correlation Matrix
CCI
Price
Negative_Social_Media
Positive_Social_Media
Neutral_Social_Media
Radio
Print
Retail
CCI
Price
Negative
SM
Positive
SM
Neutral
SM
Radio
Retail
Outdoor
Internet
TV
1.00
0.77
0.32
0.44
0.44
0.32
0.19
0.19
0.21
0.56
0.24
1.00
0.17
0.28
0.29
0.22
0.22
0.03
0.11
0.31
0.13
1.00
0.49
0.89
0.38
0.00
0.03
0.04
0.27
0.12
1.00
0.51
0.57
0.19
0.10
0.10
0.16
0.13
1.00
0.31
0.08
0.11
0.02
0.34
0.23
1.00
0.35
0.22
0.06
0.08
0.07
1.00
0.51
0.06
0.41
0.06
1.00
0.06
0.28
0.19
1.00
0.09
0.13
1.00
0.20
Outdoor
Internet
TV
0.65
Note: The variables correlation matrix indicated no major multicollinearity problems for the estimation of the parameters of the response function.
Table 4
Key Statistical Results of
theSelected Estimation of the
Current Model
R-square
93%
Adjusted R-square
90%
DW
1.85
Normality p-value
0.06
Q-stat p-value
0.05
mation approach.
level.
in Table 5.
estimation
of
multiple
parameters:
Table 5
Model Results and Statistics for Each Variable
Contri Contri
bution bution
Lambda
p-value
(%)
Omega (000) (%)
Base
VIF
1,383
51.9
0.002
1.2
Seasonality
84
3.2
<0.0001
1.4
19
0.7
0.002
3.5
49
1.8
0.004
4.0
Price
Lag
20
0.2
184
6.9
0.050
2.1
20
0.8
186
7.0
0.018
1.4
10
0.8
14
0.5
<0.0001
1.9
Radio
50
0.5
153
5.7
0.049
2.2
0.8
184
6.9
0.001
2.7
Retail
0.8
307
11.5
0.045
2.1
Outdoor
0.2
71
2.7
0.016
1.1
Internet
20
0.8
311
11.6
0.050
2.0
TV
60
0.8
188
7.0
0.016
2.3
442 JOURNAL
Note: Results of the model in terms of the percentage contribution of each explanatory variable to the modeled unit product
sales, together with p-values of each explanatory variable b coefficient estimate.
the model.
Table 6
SMR Market Pressures and Incremental Sales per Unit of SMR
Relative to the Positive SMR Impact
Positive SM
Neutral SM
Negative SM
14,762
4,457
3,320
7.0%
0.5%
6.9%
0.47%
0.12%
2.08%
1.00
0.25
4.42
In this study, the brand owners communication of these features was a critical
part of the appeal of the new technology
by consumers. When consumers still did
being
incremental-communications-
driven volume.
Only after some time was the base
Positive
Neutral
Negative
Time
Table 7
Model Results with Social-Media Variables Removed
Lambda
(%)
Omega
Type
Contri
bution
Lag (000)
Base
Contri
bution
(%)
VIF
1,632
51.7
0.000
1.3
Seasonality
83
2.6
<0.0001
1.4
38
1.2
0.000
3.5
45
1.4
0.001
2.5
Price
Radio
50
0.5
118
3.7
0.050
2.3
0.8
168
5.3
0.005
2.3
Retail
0.8
274
8.7
0.030
1.9
Outdoor
0.2
65
2.0
0.050
1.1
Internet
20
0.8
403
12.8
0.020
1.7
TV
70
0.6
422
13.4
0.023
3.5
Note: R-square 85%; Adjusted R-square 83%; DW 1.63; Normality p-value 0.01; Q-stat p-value 0.05
previously had used this service. The marketing investments variables are measured
as follows:
Television: Weekly GRP in television
advertising (Source: media agency);
Print: Weekly impressions in print
advertising:
general
content
maga-
agency);
tal agency);
model incomplete.
Weekly
promotions:
APPLICATION 2: BROADBAND
SERVICES
internal records).
sentiment
60 weekly observations.
444 JOURNAL
Table 8
Descriptive Statistics of Model Variables; Broadband Application
Mean
Acquisitions
Minimum
Maximum
Standard
Deviation
application was used to estimate the relationship between sales and independent
variables (Equation 1), and provide key
statistical results of the selected estimation
of the model (Table 10).
12,904
9,867
24,017
2,465
85
263
68
3.88
18.01
3.88
Display
96.9
436.1
107.9
Affiliates
8.53
38.54
10.86
18.44
15.99
23.41
1.97
TV
Price
Promotions
0.02
1.00
0.13
299
120
520
107
818
160
2,035
472
1,097
280
1,720
287
Note: Social-media variables (positive, neutral, and negative) are Type 3 SMR calculated according to the methodology
presented in the previous section.
for the explanatory variables had statistical significance at a reasonable significance level.
Table 10
Key Model Statistics:
Broadband Application
R-square
93%
Adjusted R-square
90%
major
DW
2.01
(Table 9).
Normality p-value
0.05
Q-stat p-value
0.05
correlations
surfaced
between
Table 9
The Variables Correlation Matrix, Indicating No Major Multicollinearity Problems for the Estimation
of the Parameters of the Response Function
Acquisitions
TV
Print
Display
Affiliates
Price
Promotions
Positive Social Media
Neutral
Social
Media
Negative
Social
Media
Affiliates
Price
Promotions
Positive
Social
Media
0.29 0.27
0.48
0.18
0.19
0.28
0.26
0.14
0.24 0.20
0.36
0.06 0.10
0.19
0.10
0.04
1.00 0.03
0.06
0.10
0.05
0.19
0.11
0.18
1.00
0.63
0.32 0.05
0.10
0.33
0.37
1.00
0.55
0.08
0.34
0.22
0.11
1.00
0.02
0.14
0.15
0.11
1.00
0.08
0.12
0.15
1.00
0.26
0.01
1.00
0.01
Acquisitions
TV
1.00
0.43
1.00
Display
1.00
Note: The variables correlation matrix, indicating no major multicollinearity problems for the estimation of the parameters of the response function.
Table 11
Model Results and Statistics for Each Variable; Broadband
Application
Lambda
(%)
Contribution Contri
(000)
bution (%)
p-value
Base
235
29.1
<0.0001
Seasonality
13
1.6
0.0052
Price
230
28.5
<0.0001
Omega
Lag
Promotion
0.5
16
2.0
0.0017
35
0.3
82
10.2
0.0152
TV
87
0.7
197
24.4
0.0004
45
0.7
58
7.2
0.0511
Online Display
10
0.3
55
6.8
0.0004
Online Affiliates
0.4
111
13.8
<0.0001
Note: Results of the model in terms of the percentage contribution of each explanatory variable to the modelled unit product
sales, together with p-values of each explanatory variable b coefficient estimate.
modity and does not command high customer loyalty levels. The results of the
Without
social-media
variables,
the
because, over the authors period of observation, prices came down, thereby having
a positive effect on the new subscriptions
and renewals.
Of the tested three types of social-media
messages, only the negative-sentiment
Table 12
Model Results with Social-Media Variables Removed; Broadband
Application
Lambda
(%)
Contribution Contri
(000)
bution (%)
p-value
Base
220
27.3
<0.0001
Seasonality
11
1.4
0.0259
Price
185
22.9
<0.0001
Omega
Lag
Promotion
0.5
15
1.8
0.0022
TV
88
0.7
186
23.1
0.0028
45
0.7
60
7.5
0.0810
Online Display
12
0.3
49
6.1
0.0019
Online Affiliates
0.5
102
12.6
<0.0001
446 JOURNAL
Note: R-square 91%; Adjusted R-square 87%; DW 1.79; Normality p-value 0.07; Q-stat p-value 0.06
authors
observed
that
negative
MANAGERIAL IMPLICATIONS
marketing.
concentrated.
social-media
one-way
the industry;
business outcomes.
different
Internet,
outdoor,
retail,
capability.
customer
stages.
In
traditional
communication,
With
social
media,
and
approvals.
environments.
communications.
versation topics.
current study:
the-art
should
be
wrong
classifications.
linguistic
used
to
algorithms
minimize
such
effect on sales.
social media.
be analyzed carefully.
multiple industries.
448 JOURNAL
References
2010.
935975.
(2003): 111.
tagunta .
1 (2011): 224238.
2005.
Copyright of Journal of Advertising Research is the property of Warc LTD and its content may not be copied or
emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.