Our social media measurement approach applied to a popular personal care brand. Here we assess the degree to which PAID media is AMPLIFIED by EARNED media conversations.
2. Brand: Brand X
Case study authors: Masood Akhtar & Michael Wolfe, Bottom Line Analytics Global
Period of case: Jan 2012 to April 2014 (Analysis undertaken in June/July 2014)
Case Details
3. Background
Personal care brand X has seen very little sales growth year on year. Many of the recent and on-going marketing
and media campaigns have and continually try to ignite the brands intrinsic qualities for its’ target audience of
females aged between 24 – 39.
The marketing mix still includes many of the traditional offline channels, although in recent times this has shifted to
include more paid digital and social media.
What was the Campaign/Activity designed to do?
This was a classic campaign designed to make consumers appreciate the intrinsic features of the brand.
Why was social media used?
Social Media Display ads on Facebook were used to illustrate, reinforce and engage the target consumer to the core
messages around the intrinsic product features.
Market Context for Personal Care Brand X
4. Project Objectives
1. Can we measure the effectiveness (contribution and headline ROI) of both offline and online media (including
paid social media)?
2. We recognise that social media has both a paid and earned dimension (voice of the customer). The latter is
becoming increasingly important for brand X and its’ competitors, so can we measure any real effect of this on
retail sales volume?
3. With the media landscape being so fragmented, can we assess the interactions between media and broader
social media voice of the customer to determine a synergistic impact on sales – this will help with future media
execution strategy.
4. Optimise media spending across all channels to increase sales volumes for personal care brand X.
5. Fusing SEITM based language
measurement with
advanced analytics to
understand competitive
brand positioning, content
drivers and reputation.
Using known tools to listen
and monitor high level
consumer conversations.
Measure language based
on engagement and
importance through the
Semantic Engagement
Index (SEITM).
Listening,
Monitoring and
basic Sentiment
Measuring
Language for
brand insights
Social
Media Advanced
Analytics
Social
Monetization
Applying a trended SEITM
within Media Mix
Modelling to monetise
Social Voice of the
Customer (earned social
media) alongside all other
media and quantify any
synergistic effects.
Extend the Value of Social Media Insights
BLA Social Insights, Analytics and ROI Framework
6. About Us
The Semantic Engagement Index SEITM is a product of Stance-Shift AnalysisTM*. Published and peer
reviewed, Stance Shift AnalysisTM reveals what really matters to the consumer. This is the underpinning of
our approach. Stance-Shift measures consumers’ verbal shifts in positioning as they talk, where the “shift”
infers a landmark change in emotion, intensity, appraisal and commitment towards a subject (brand,
topic, campaign or concept)
This approach enables us to solve for what others miss: Size, Trend and New Concepts.
Two levels of scoring:
1) Emotion and Commitment is understood through engagement scoring – far superior to simple
words/comment frequency.
2) Graded tonality scoring (the + and -) to understand negative and positive sentiment.
The Semantic Engagement Index SEI
TM
integrates our Stance Shift measurement engine to power
consumer insights and advanced analytical modeling.
Using Stance Shift AnalysisTM for consumer insights
and modelling
Stance Analysis: Social cues and attitudes in online interaction, Mason, et al, Linguistic Insights
7. Decoding Text
Example of an Experiential Comment
I just got my cool new iPhone from BestBuy,
However, I keep getting dropped calls on the
Brand X 4G network
Typical ‘Sentiment’ metrics are based on automated word counting and binary scoring.
The comment above would be counted for positive and negative which in many cases would
ultimately be bucketed into neutral overall.
Positive
Negative
8. I just got my cool new iPhone from BestBuy,
however, I keep getting dropped calls on the
Brand X 4G network
Positive
Negative
Flag Brands & Relative Importance
Custom coding
Engagement
8
1. Evaluate the Entire Conversation
2. Account for Context
3. Adapt to Industry Language, Terms
4. Adjust to Channel Communication (Facebook, Twitter, specialist forums, blogs)
Decoding Text using Stance Shift AnalysisTM
Teasing out the nuances of language
Transitional
word (Shift in
Stance)
9. From Millions of Cleaned
social media
Conversations
Powerful social insights on topics
and themes of conversation that
are most engaging.
Small Pepermint Afternoon Snack 12Pack
Great Deal Breakfast yum Large
Miss it Get me one Orange on sale
Morning Half Priced got coupon Drive Home
Vanilla Mocha 8Oz need a hit
We Detect Thousands of interesting
“nodes” of Consumer information
Clear Topics &
Concepts of
Importance
Emerge
We apply advanced
Analytics to help drive content
strategy and measure social ROI.
Our Supervised Learning Pattern
Detection organizes the nodes
The Solution Path For Consumer Chaos
10. Developing the personal care brand X semantic engagement
metric
1. Mine social media
conversations and online
reviews/blogs for Personal Care
Brand X and competitor brands.
2. Apply our Linguistics based
engagement* and tonality scoring
to textual corpus.
3. Develop a weekly consumer semantic
engagement metric for brand X and 4
competing brands.
• Here we define ‘Engagement’ as it transpires from the use of language on social media.
11. Modelling Design: Two stage modelling
Digital Display banners
Personal care Brand X
Semantic Engagement metric
Paid Search
Mobile Ads
Brand X competitor Eng. metric
OOH Media
Radio
Personal Care
brand X
Nielsen Retail
Sales VOLUME
by Week
TV
Print Media
Cinema Ads
Retail Distribution
Retail Price
Digital
Semantic
Engagement
metric
Traditional
Offline
Mass
Media
Paid Social Media Display Ads
Recurring market seasonality
Retail
Key data sources: Nielsen, Brandwatch, IPSOS and client databases. Period of case: Jan 2012 to April 2014 (Analysis undertaken in June/July 2014)
Earned Media
12. How the data were analysed.
A comprehensive time-based marketing response model was developed covering all paid traditional and digital
channels, plus earned media effectiveness due to social media brand conversations covering all social channels. In
addition, our model included earned media from 4 key competitors which illustrated and measured the negative
impact from the social engagement of competitive brands on Personal Care brand X. Incremental volume gains
from the beginning of the campaign were up 28%, 19% and 26%, respectively from earned social media, traditional
media and paid digital media versus the prior year.
13. A note on Causality in Social Media
Causation is near impossible to prove. However, we evaluated our social media engagement metric (earned social
representing the voice of the customer) to determine if it was a leading, coincident or lagging indicator of brand
retail sales volume. Our findings reveal that the highest time based correlation of our social media engagement
metric was 1 week prior to sales. This indicated that social was a leading indicator of sales. As a leading indicator,
therefore, this supports the notion that positively engaged consumer chatter on social media acts as a predictor of
sales rather than an artefact of sales.
14. Social Voice of the Customer correlates with retail sales
The brand semantic engagement metric captures the voice of the customer as it is reflected through the “customer
experience” in social media brand conversations.
Correl.= 51%
The unadjusted engagement metric has a lead time of
approximately 1 week on retail sales. It takes positively engaged
social media commentary for brand X a week to impact sales. The
correlation in week 1 is also statistically significant at the 0.05%
level, which means that if this were to be repeated with similar
data this relationship may well hold true 95/100 times.
For our Personal Care brand X, this shows a strong
51% correlation to retail sales volumes which means
there is a strong association between positively
engaged social media commentary (as represented by
SEI) and sales. The impact is through an
amplification of other effective media channels.
15. Personal Care Brand X Modelling Accuracy
R2 = 92.9%, Holdout R2=93.3%, MAPE= +/- 1.3%
Out of sample holdout
forecastThis is the period we modelled. Our model explains 92.9% of the
movements in actual retail sales volume.
Our out of
sample holdout
test indicates the
‘predictive’
capacity of this
model – at 93%
this is very high.
16. Contribution in the last 12
months to 4/26/2014
Percent
Annual
Contribution
Base 106.9%
Personal care brand SEI 3.6%
Competitor brand SEI -5.8%
Digital Media 2.0%
Offline Media 3.8%
Retail Price -12.2%
Retail Distribution 1.8%
Modelled Retail Sales Contributions for brand X
Baseline Sales Volume
Positive consumer engagement across social media for
the personal care brand X has contributed 3.6% of
sales. Brand X campaigns will be extended as a result of
this.
Competitor consumer engagement has a greater
negative impact (-5.8%) on brand X’s sales.
Baseline sales includes
previous accumulations
of brand equity and
periodic seasonality.
Consumer engagement for competitor
brands is having a greater impact on
brand X’s sales than the positive impact
brand X’s own engagement has
generated.
Baseline Sales for
brand X
About 6% of sales are due to digital and mass media
17. Mobile media, Display ads, Print and Radio together drive the highest revenue per investment for our Personal care brand X.
Note, the brand Semantic customer engagement metric is not represented here as it is not a media vehicle. Rather, it is the
voice of the social customer – a feedback mechanism that that amplifies the impact of other channels.
Marketing and Media ROI
Note:
1) Above numbers for the last 12 months only.
2) No margin data was provided, so ROI at Profit was not calculated.
3) TV was ineffective largely because of the creative messages – scores for message liking and recall were particularly low.
4) Paid social media banners appear to show a loss in this study – this could vary depending on the social platform used as often they can provide great reach and
brand awareness.
5) No in-store activity/spend in the last 12 months, so no contribution.
6) In-store features and display data was not provided so was not modelled at all.
Most productive channels
Least productive channels
Media Channel
Revenue
Contribution
USD
Current Spend
(Investment)
USD
Revenue /
Investment
(ROI)
Print $16,294,087 $1,424,347 $11.44
Radio $3,563,403 $410,728 $8.68
Online Display Banner $9,890,101 $1,290,860 $7.66
Mobile $2,119,698 $319,781 $6.63
TV $3,798,921 $15,430,180 $0.25
Paid Social Media Display Banners $14,590 $406,328 $0.04
Online Paid Search $15,185 $585,288 $0.03
Out of Home $19,743 $1,945,092 $0.01
Cinema $6,050 $862,529 $0.01
In Store $0 $0 -
Total $35,721,778 $22,675,133 $1.58
18. The Retail Sales Contributions from earned
social media (as represented by the SEITM)
The Semantic Engagement Index has contributed 3.6% of the incremental sales revenue. This metric has no spend component which makes
it impossible for us to apply the usual ROI formulae. However, as the SEI represents the social voice of customer it amplifies other effective
media channels. The impact of this amplification is shown below.
The largest amplification impact of
SEI was for Mobile and online
display banners.
19. Digital and mass offline media sales response
The chart below shows sales response for both Offline Mass Media (mostly TV) and Digital Media, mostly paid search
and Display banner ads). The bending mass media curve illustrates the diminishing returns dilemma for Mass offline
Media, while Digital Media does not have this problem at current levels.
20. Optimising total Personal Care Brand marketing spend
The premise of this exercise is to reallocate spend from less effective channels to the most effective channels such that we see an uplift in
sales. This is known as a full spend optimisation which holds the marketing budget constant and assumes that all else is equal in terms of
message quality and creative strength.
Model shows +12% sales gain at constant spend
Revenue Contribution $ Current Spend $ Optimal Spend $
Cinema $6,050 $862,529 $172,506
Print $16,294,087 $1,424,347 $4,273,041
In Store $- $- $-
TV $3,798,921 $15,430,180 $11,578,136
Radio $3,563,403 $410,728 $1,232,185
Out of Home $19,743 $1,945,092 $389,018
Social Media $14,590 $406,328 $81,266
Online Search $15,185 $585,288 $117,058
Online Display $9,890,101 $1,290,860 $3,872,581
Mobile $2,119,698 $319,781 $959,344
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Cinema
Print
In Store
TV
Radio
Out of Home
Social Media
Online Search
Online Display
Mobile
Spending optimization
calls for nearly doubling
digital spend!
21. Measuring media synergies: Personal care Brand X
semantic engagement & all media
Our analysis found that the key drivers behind brand X’s customer engagement were the intrinsic elements of the
product proposition. Driving up the brand X’s semantic customer engagement score therefore not require spend, but
for media communications vehicles to carry messages that would trigger positively engaged conversations around the
intrinsic benefits of brand X.
2.2%
Extra
$1.23
The chart on the left shows the independent incremental sales impact of the customer semantic engagement metric
and all media for brand X, as each are executed separately and independently (left bar). When each of these are
executed concurrently, there is a dividend or synergy where the total is greater than the sum of the independent
parts. The right chart shows the incremental ROI due to this synergy. This synergy is a measure of the core benefit of
integrated marketing.
Synergy
22. About 6% of Personal Care brand X sales are due to digital and mass media.
Positive social engagement for Personal Care brand X (vis-a-vis our brand semantic engagement metric) has generated
3.6% of sales.
Across the media mix, Digital Display ($7.66), Print ($11.44), Radio ($8.68) and Mobile ads ($6.63) were found to
generate the highest returns per Dollar spent. By contrast, TV ($0.25), Cinema ($0.01), OOH ($0.01), Online Search
($0.03) and Social Media ads ($0.04) generated comparatively weak returns.
Personal Care brand X can double up spend on Digital without fear of saturation and diminishing returns. Offline
media is showing signs of saturation, yet it is vital to integrate offline with digital more due to the synergistic impact on
sales.
Positive consumer engagement on social media is also crucial in terms of integration with all media. Driving up
consumer engagement across social media would require a concerted effort to develop communications (channeled
through offline and online media vehicles) that trigger important topics and the amplification via earned social. In the
case of brand X, those conversations would be related to intrinsic product features.
Our optimized media spend solution with the same budget would lead to a 12% uplift in sales volumes.
Modelling Insights Summary
23. Main lessons for Personal Care brand X
One of the key objectives of the modelling exercise was to demonstrate the efficacy and role of earned social media
on business performance for the brand. To this end, the project was very successful to the extent that (1) we
demonstrated the impact of Brand X and competitor social conversations on brand purchases, (2) what particular
topical content was most relevant in forming the engagement metric driving brand purchases (this is not included
due client confidentiality request) and (3) that earned social media was a lower-purchase-funnel metric with a
direct impact on purchases rather than an upper-funnel metric with only an indirect effect through its impact on
brand awareness. Our Structural Equation Modelling demonstrates this.
What we would have done differently?
Applied the Paid, Owned, earned framework more explicitly and report results in this way.
24. Analysis of the broader insight eco-system to understand the rightful place of social media.
In addition to marketing econometrics model, we also constructed a Structural Equation Model (SEM). The goal of
this model was to better understand the blueprint of how marketing works. This model included retail sales, our
social customer engagement metric, survey-based brand awareness and brand perceptions data, plus offline and
digital media data.
The key insight of this model was to demonstrate that social media voice of the customer was a lower-purchase-
funnel metric rather than an upper funnel metric. That means that social brand conversations have more of a direct
impact on purchases rather than an indirect impact by affecting brand awareness at the upper funnel.
25.
26. The Role of semantic customer engagement within the
Personal care brand X Insight Ecosystem
Key Insights
1. Offline media transmits through to positive consumer chatter via awareness channels.
2. Digital display is strongly associated with consumer engaged volume for Personal care brand X on social
media.
3. Positive social chatter about Personal care brand X is strongly linked with agreement on intrinsic Personal care
brand X attributes (including likelihood to recommend) from survey based tracking.
4. The positive consumer engagement metric is positioned closer to retail sales as opposed to social media
display which is more of an upper funnel metric, driving brand awareness.
27. Michael Wolfe
CEO
Bottom Line Analytics (US)
E: mjw@bottomlineanalytics.com
M: 770.485.0270
www.bottomlineanalytics.com
Masood Akhtar
Managing Partner
Bottom Line Analytics EMEA
E: ma@bottomlineanalytics.com
M: +44 7970 789 663
www.bottomlineanalytics.com