Enterprise Competitive Analysis and

Consumer Sentiments on Social Media

Insights from Telecommunication Companies

Eric Afful-Dadzie

, Stephen Nabareseh

, Zuzana Komínková

Oplatková

and Petr Klímek

Faculty of Applied Informatics, Tomas Bata University, T.G Masaryka 5555, 760 01, Zlin, Czech Republic

Faculty of Management and Economics, Tomas Bata University, T.G Masaryka 5555, 760 01, Zlin, Czech Republic

Keywords: Sentiment Analysis, Competitive Analysis, Telecommunication Companies, Social Media, Facebook,

Twitter.

Abstract: The utilization of Social media tools in business enterprises has tremendously increased with an increased

number of users and a corresponding upsurge in time spent online. Online social media services such as

Facebook and Twitter are used by companies to introduce new products and services, provide various

supports and interact with customers on daily basis. This regular interaction of businesses and consumers

results in huge amount of customer-generated content which is becoming a source of insight in analysing

the often erratic consumer behaviour. For companies to harness the business potential of social media to

increase competitive advantage, sentiments behind textual data of both their customers and that of their

competitors must be keenly monitored and analysed. This paper demonstrates how companies especially

those in the Telecommunication industry can seize the opportunity presented by social media to mine

textual data to gain advantage over competitors by cumulatively understanding consumer opinions,

frustrations and satisfaction. Using Facebook and Twitter sites of the top three telecommunication

companies in Ghana: MTN, Vodafone and Tigo the paper reveals insights from unstructured texts of

customers of these three companies. The results show (1) the exponential growth of social media users in

Ghana (2) impact and numbers behind active social media participation in the telecommunication industry

(3) the power of social media opinion mining for competitive analysis (4) how business value could be

extracted from the huge unstructured textual data available on social media and (5) the company that is

more responsive to customer concerns.

1 INTRODUCTION

Online social media is now an integral part of our

lives. Recent research works by (Wollan and Smith,

2010; Barlow and Thomas, 2011; Qualman, 2009;

Safko, 2010) corroborate the exponential growth of

social media as a new strategic asset for businesses.

In particular, (Barker, 2008; Sinderen and Almeida,

2011; Weber, 2009; Gillin and Schwartzman, 2011)

enumerate some of the ways social media can be put

to use by businesses. Some of these are (1)

identifying new product ideas (2) finding new

business opportunities (3) creating brand awareness

(4) strengthening customer relationships and (5)

establishing contacts with employees, partners and

even with competitors. A recent study report by

McKinsey (McKinsey, 2013) on how organizations

are using social media tools also reveals a range of

benefits for business enterprises as shown in table 1.

For example 69% of the respondents reported an

increase in how effective marketing on social media

has been to their companies with 52% reporting an

increase in customer satisfaction. Despite these

purported benefits, Harvard Business Review (HBR,

2010) posits that though companies are aware of the

huge business potential of social media, most of

them are only making investments for the future

because they still doubt or view the potential as

hyped.

Even as the debate on the potential of social

media to businesses lingers on with little work from

academia to support the discussion according to

Jussila et al. (2014), the wealth of textual data from

social media continues to pile up on daily basis.

Textual data especially those expressing

concerns, frustrations and acceptance from

Afful-Dadzie E., Nabareseh S., Komínková Oplatková Z. and Klímek P..

Enterprise Competitive Analysis and Consumer Sentiments on Social Media - Insights from Telecommunication Companies.

DOI: 10.5220/0004991300220032

In Proceedings of 3rd International Conference on Data Management Technologies and Applications (DATA-2014), pages 22-32

ISBN: 978-989-758-035-2

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

Table 1: Business and Web 2.0 - Benefits from customer

use to organizations. Source: McKinsey.

Benefits (%)

Increasing effectiveness of marketing 69

Reducing marketing costs 47

Reducing customer support costs 36

Reducing travel costs 47

Increasing customer satisfaction 52

Increasing revenue 24

Increasing number of successful innovations 22

Reducing time to markets 28

customers are rich in knowledge which needs to be

mined for insights. The wealth of knowledge behind

customers’ comments, posts and tweets on “wall

pages” of companies could prove valuable in

identifying popular brands and level of customer

satisfaction among others (Bradley and McDonald,

2011; Governatori and Iannella, 2011). For example

the ‘‘like” feature on Facebook which allows users

to approve a post can prove valuable in identifying

types of users and their interests in a service or

product (Lipsman et al., 2012).

The emergence of online social media now gives

consumers a rare but powerful influence for publicly

expressing their opinion and thoughts on products

and services. This brings revolution in the way

customers and companies interact by proffering new

ideas about customer relationship and brand

management. To process, aggregate and gauge the

influence of customers’ opinions, sentiment analysis

and text mining in general are used. Feldman (2013)

explains the role of sentiment analysis in

organizations and how it could be employed to

monitor the activities of customers in the various

social media sites in real time.

In the following sections, we briefly explained

the concept of sentiment analysis and its related

applications, followed by research questions posed

for this paper. Next is the methodology including the

samples and procedures used. A case study of the

sentiment analysis approach is conducted using the

top three telecommunication companies in Ghana.

The key findings are discussed followed by how

responsive the companies are to customer specific

concerns.

2 SENTIMENT ANALYSIS

Sentiment Analysis sometimes referred to as opinion

mining often combines the study of Natural

Language Processing (NLP), Statistics and Machine

learning to extract and classify the sentiments of a

textual input. There exists two basic sentiment

analysis tasks namely subjectivity and polarity

detection. In the subjectivity approach, prediction is

made on whether an inputted text is subjective or

otherwise whiles the polarity technique makes an

overall prediction of whether a subjective text is

positive, neutral or negative (Buckley and Paltoglou,

2012). The polarity technique learns to classify the

polarity (positive or negative) of a statement,

comment or post and determine whether the

sentiment behind a statement or comment is positive,

negative or neutral (Koppel and Schler, 2006; Pang

and Lee, 2008; Dave et al., 2003; Pang et al., 2002;

Wiebe, 1994). In the process a document is flagged

as positive, negative or neutral. A score of zero

denotes a neutral sentiment.

Sentiment analysis, unlike classical text mining

which focusses on topical words, picks only

sentiment signals for real time analysis (Pang and

Lee, 2008).

Additionally, two main approaches are used in

the task of automatically extracting sentiments from

text. These are text classification and the lexicon-

based method. In the former approach, a classifier is

typically built from labeled instances of texts or

sentences in a supervised classification

(Turney,

2002). This employs the use of statistical and

machine learning techniques. Some of the most

popular algorithms and techniques employed in text

classification based sentiment analysis include but

are not limited to support vector machines, naive

Bayes, Ada Boost, k-nearest neighbours, and

maximum entropy (Pang et al., 2002; Cao et al.,

2013). These techniques are used when appropriate

with Natural language processing and statistical

techniques.

On the other hand, the lexicon-based approach

works by computing the polarity of a document

based on the semantic orientation of words or

phrases in the document (Toboada et al, 2011;

Vaithyanathan, 2002) In practice, the lexicon-based

approach depends on dictionaries of words marked

with the word’s semantic polarity in computing the

overall polarity of a document. Some of the widely

used dictionaries for lexicon-based approaches are

SentiWordNet and Wordnet though a customized

dictionary can be created (Tong, 2001) using seed

words to expand the list of existing words as

demonstrated by Taboada et al (2011). In this paper,

we use the lexicon-based approach with the

SentiWordNet (Esuli and Sebastiani, 2006)

dictionary to generate the sentiment scores.

Sentiment analysis has been applied to a wide

range of domains and disciplines. Notable among

these are in the area of consumer product and

EnterpriseCompetitiveAnalysisandConsumerSentimentsonSocialMedia-InsightsfromTelecommunication

Companies

services reviews (Feldman, 2013) and in voter

sentiments in politics (Pang and Lee, 2008). For

example (Zhang and Wan, 2013) used sentiment

analysis to find possible weaknesses of products

from customers’ feedback. Kang and Park (2014)

designed a new framework that uses sentiment

analysis and VIKOR method to measure customer

satisfaction in a mobile service. Ghiassi, Skinner and

Zimbra (2013) also combined neural network with n-

ngram for twitter sentiment analysis and to show that

their proposed Twitter-specific lexicon is

significantly more effective in classification recall

and accuracy metrics than some of the traditional

twitter lexicons. A sentiment analysis based on

unsupervised and domain-independent model was

and implicit aspects in reviews whiles (Bai, 2011)

proposed a heuristic search-enhanced Markov

blanket model to predict consumer sentiments from

online text.

The interest in sentiment analysis has brought

about a wide range of tools both commercial and

open source. Dyer (2013) lists some of the top 50

sentiment analysis tools. These tools help find

insights and patterns from unstructured textual data

(Sashi, 2011; Zeng et al., 2012b).

This paper used a python programming language

module called pattern together with SentiWordNet

to generate the sentiment scores of user comments

for the research. In the following sections we explain

the sentiment analysis architecture adopted, the

methodology guiding the research and subsequently

use a case study of consumer comments of the top 3

Telecommunication companies in Ghana. The

resulting sentiment scores are then used for a

competitive analysis.

3.1 Research Questions

Studies have shown a gradual interest in user-

generated content on social media (Feldman, 2013;

Akehurst, 2009; Aggarwal et al., 2011). This paper

employed opinion mining to do a competitive

analysis using unstructured textual information on

Facebook and Twitter sites of the top 3

telecommunications company. The data was

collected from 02-02-2014 to 07-03-2014. The

following research questions guided the study.

1. What patterns can be found from their

Facebook and twitter sites respectively?

2. How do customers feel about their products

or brand?

3. Which Telecom company is most

competitive as far as issues of customers are

concerned?

3.2 Methodology

The study was carried out in phases to help answer

the research questions. First, descriptive quantitative

data was collected from the respective individual

social media sites. In particular, we collected the

number of tweets, followers and following from

Twitter and the likes, talking about this and notes

features from Facebook. Relevant for the research

were also the demographics and geographical

location of users, number of postings from the

companies, comments made by users, time of the

day when most posts are made and the frequency of

postings between 02-2014 and 02-03-2014 as shown

in figures 2-6 respectively. Next, we applied

sentiment analysis techniques to analyse and flag

each user comment as negative, positive or neutral.

This helped in discovering the sentiments behind

each user comment, the manner and pace each

company responds to user concerns and the general

pattern of how the three companies have been active

on social media in relation to their customers.

The sentiment analysis of user comments on

social media adopted for the competitive analysis of

telecom companies follows the architecture shown in

figure 1. The methodology adopted is basically

broken into three steps below.

Step1: Text pre-processing:

In the pre-processing stage, we first used Repustate -

an online analytical tool to ‘scrape’ user comments

from the social media sites (Facebook and Twitter)

of the 3 top telecommunication companies. Each

user comment, post or review extracted was treated

as a document. The collection of documents-the

corpus was saved in comma separated value format

(CSV). Then the documents in the corpus were each

converted to text and pre-processed using linguistic

tools basically of tokenization and stemming in

python pattern module. The pre-processed corpus

was loaded together with the SentiWordNet

dictionary into python for subsequent processing.

Step2: Text processing

The sentiment analysis approach adopted for this

research is the lexicon-based approach. First, each

document (e.g., post, tweet, review etc.) collected is

tokenized into a word list. Next each token’s log

probability is identified in the word list. The log

probabilities of each token are then added together to

determine the probability of each sentiment for the

entire document. To determine a binary ‘positive’ or

‘negative’ label with scores for each document in the

corpus (user comment), the SentiWordNet scores are

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

accumulated with the more subjective comment

having a higher score.

Figure 1: Lexicon-based sentiment analysis architecture

for social media.

Step3: Python pattern vrs other software tools

To ascertain the effectiveness of the sentiment

analysis techniques we adopted and subsequently the

polarity scores for each user comment, comparison

was made with other sentiment analysis tools

notably RapidMiner and Python NLTK 2.0.4 text

classification. The RapidMiner sentiment analysis

tool detected the polarity of a document without

sentiment scores. However the results turned out to

be consistent with our results in terms of the polarity

(negative, positive or neutral) assigned to each of our

user comments. In the case of Python NLTK text

classification, whiles almost all the polarity labels

of ‘positive’, ‘negative’ or ‘neutral’ label for each of

our documents were consistent, the scores assigned

to each user comment turned out to be slightly

different. For instance in Table 5 row 3, the

comment “

MTN not good at all” though a negative

polarity in both Python pattern and NLTK text

classification, was assigned a score of -0.6 in NLTK

compared to a -1 in Python pattern module.

3.3 Case Study

The Telecommunications Industry in Africa

continues to grow at an unprecedented rate

(McKinsey, 2010; Chavula, 2013; Mahmoud and

Hinson, 2012). In Ghana, there are currently six (6)

major operators providing wide range of services

such as cell phone telephony, broadband internet,

and mobile financial services among others. These

are as at 2013, Expresso, Millicom Ghana known in

the market as Tigo SCANCOM known as MTN,

Vodafone Mobile, Airtel, and Glo Mobile. According

to the National Communications Authority (NCA); a

body mandated by the Government of Ghana to

among other things, grant licenses, monitor telecom

service quality and protect consumers; Ghana’s total

mobile subscribership currently stands at 28,026482,

a growth of 7.4% from 2012 (NCA-Ghana, 2013). In

spite of the apparent growth, service quality

continues to be a challenge for consumers prompting

the NCA-GH to slam five (5) of the operators a total

fine of USD 464,413.50 in 2011. The advent of

social media has therefore provided consumers in

Ghana, a platform to interact with telecom providers

in real-time; venting frustrations, reviewing products

and generally receiving customer supports.

It is on this premise that we use sentiment

analysis to find how telecom companies in Ghana

competitively respond to issues of customer

concerns and also generally delve into what

consumers think of these companies. We use the top

three telecom companies (by subscriber base) in

Ghana according to (NCA-Ghana, 2013). These are

MTN, Vodafone and Tigo respectively in the order of

customer size. As at December 2013, MTN had a

subscriber base of 12,929,528 users with Vodafone

and Tigo recording 6,048,792 and 4,021,225

respectively.

The selection of the three (3) telecommunication

companies for this research is necessitated by their

strong presence on social media compared to the

other operators. MTN, Vodafone and Tigo interact

with their customers almost on daily basis The time

period of Feb 2 –March 7 was also chosen because

of two important events within this period. These

events are Valentine’s and Independence Day

celebrations. During these events, a lot of telephony

activity is recorded because of the large numbers of

people both youth and elderly who access

telecommunication services. Our initial findings in

this research reveal that MTN made an average of 1

and 9 posts on their Facebook and Twitter pages

respectively between 02-02-2014 to 07-03-2014.

Vodafone made an average of 4 and 2 posts on

Facebook and Twitter respectively whiles Tigo

posted an average of 2 and 1 message respectively as

shown in table 2 and 3.

In figures 2 and 3, the daily trend of user

comments posted on Twitter and Facebook is shown.

The trend shows that within the period of the

research, MTN users interacted more with their

telecom operator on Facebook than Vodafone and

TiGo users. On Twitter, TiGo users exchanged more

Step 2: Text Analysis

Document

processing

Document

analysis

Lexical

resources

Sentiment

scores

Facebook &

Twitter

Comments extraction

Pre-Processing

Corpus

from user

comments

Step 1: Text pre-processing

EnterpriseCompetitiveAnalysisandConsumerSentimentsonSocialMedia-InsightsfromTelecommunication

Companies

Figure 2: Trend of Tweets from Feb 2 - March 7, 2014 for the 3 companies.

Figure 3: Trend of Facebook posts from Feb 2 - March 7, 2014 for the 3 companies.

comments and posts between them and their

operators than as it happened on Vodafone and

MTN.

3.4 Preliminary Findings

The web scraping for textual data (posts, comments,

tweets etc.) returned valuable information about the

social media activity of the three telecom

companies. This included the number of posts made

in a day, number of likes/followers, and other

features specific to one social media such as notes,

talking about this on Facebook and following on

Twitter as shown in tables 2 and 3 respectively.

The number of tweets and posts on Twitter and

Table 2: Stats on Facebook Feb. 2 – Mar. 7, 2014.

Likes

No.of

Posts

Talking

about this

Notes

Tigo 313,287

67

6,011 13

MTN 211,649

289

3,179 19

Vodafone 274,863

127

1,600 143

Facebook respectively were generated within the

month dedicated to the research.

The study shows that the three companies seem to

have a strong presence on Facebook than on Twitter.

MTN for example posted 289 messages on their

Facebook wall compared to only 46 on Twitter in

the same period as shown in table 2 and 3. Vodafone

recorded 127 posts on Facebook with 89 on twitter

whiles Tigo managed 67 posts on Facebook and 74

on Twitter as indicated in tables 2 and 3. The trend

showed MTN as the most active telecom company

in Ghana on Facebook as far as posts emanating

from the company was concerned whiles Tigo

appeared slightly the most active on Twitter in the

period as shown in figures 2 and 3.

Table 3: Stats on Twitter from Feb 2 - Mar 2, 2014.

Tweets Followers Following

Tigo 74 9,834 6,084

MTN 46 30,170 3,147

Vodafone 89 21,911 1,743

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

Table 4: Classification of sentiments on Facebook.

Negative Positive Neutral Total

Tigo 161 381 462 1004

MTN 101 297 291 689

Vodafone 301 966 1721 2988

It can be seen in figures 2 and 3 above that the

peak time of posts and tweets on Facebook and

Twitter respectively did not occur at the same time

period. A content analysis of the entire posts and

tweets revealed that the reason for the discrepancies

was because the three telecom companies embarked

on different events, deals, special monthly offers

such as discounts and incentives at different days.

MTN for example had an SMS scam alert message

to their fans/customers on Facebook which

generated responses and re-posts within the period.

The disparity in the postings and tweets was again

evident in the times of the day that customers/fans

interacted with their telecom providers on social

media. In figures 4-6, an interesting pattern emerges

on the time of the day that customers/fans were most

active on the pages of these three companies. The

action intelligence to be extracted from this would

be the ideal time for the companies to target their

customers/ fans with new products, special offers

and important messages such as alert on scams as

was done by MTN.

4 KEY FINDINGS

This section briefly explains some of the key

findings from the social media sites of the three

telecom companies as far as business competitive

and sentiment analysis were concerned. In table 4, a

summary of the classification of all the textual

information extracted from the social media sites of

the three companies into negative, neutral and

positive comments is presented.

On Facebook, Vodafone had the largest number

of interactions than MTN and Tigo within the period

of the research. A total of 2,988 textual information

were extracted from Vodafone’s Facebook wall.

This number includes both posts originating from

the company and the responses/comments generated

as a result of the posts. The implication is that, with

127 posts on Vodafone’s Facebook’s wall as seen in

table 2, it elicited 2,861 responses from

fans/customers. Competitively, it means a far

positive customer-company interaction occurred for

Vodafone than MTN which posted 289 posts (see

table 2) but generated a paltry 654 user comments

and responses within the period. It is also worthy to

note that even with Tigo’s 67 posts on their wall

within the month of February to March, customer

generated comments/responses and posts (1004) was

more than that of MTN.

4.1 Sentiment Analysis on Facebook

The sentiment analysis architecture adopted

generated sentiment scores of -1 to +1 indicating

most negative to most positive comments to classify

the sentiments behind the textual data on the three

companies. As shown in table 4, Vodafone had 301

negative comments representing 10.07%, 966

positives (32.33%) and 1,721 (57.60%) neutral

comments out of the total of 2,988 texts generated.

The classification generated 161 negative comments

on Tigo representing (16.03%), 381 (37.95%)

positives and 462 (46.02%) neutral comments. MTN

on the other hand had 101 negative comments

representing 14.66%, 297 positives (43.10%) and

291 (42.20%) neutral comments.

Whereas a cursory analysis of table 4 does not

reveal any disturbing trend for any of the three

companies as far as the numbers recorded for the

total negative comments is concerned, content

analysis of the actual individual comments and their

relative sentiment scores shown in tables 5-7 is

telling on the performance of the companies and

their brands in Ghana.

The individual sentiments from the three companies

in tables 5-7 reveal gravely dissatisfied customers

with strong statements of dissatisfaction. For

instance comments like “

Airtime s**kers. Can't even

browse for 30mins with 2gh airtime” for MTN, “tigo is

the most useless network in Ghana. bought the ghc39.99

bundle its too slow. I can't even open common google. you

have just lost a customer and I will never” for Tigo and “a

disappointing network, in Takoradi but yet barely

able to

connect. would not recommend” for Vodafone

signify discontent. To gain competitive advantage,

the companies must respond positively and quickly

to these comments to have an edge over their

competitors

. In-depth content analyses of the

positive comments on Facebook reveal interesting

responses by the three companies to specific

customer queries or enquiries. Tables 8-10 are some

specific responses to enquiries by their customers.

Customers feel more relaxed and cared for if

they are being identified by name on specific and

direct responses to their concerns (Mittal and Lassar,

1996; Peppers and Rogers, 1995). From our content

analysis and as can be seen in table 8, MTN

EnterpriseCompetitiveAnalysisandConsumerSentimentsonSocialMedia-InsightsfromTelecommunication

Companies

Table 5: Some selected MTN negative comments.



Comment Score

1

Airtime s**kers. Can't even browse for 30mins with 2gh airtime. -1

2

Its annoying if as at now you call yourselves best network in Ghana and still we can't get good

network to make calls for browsing, don't wanna talk, I know what to do. :(

-1

3

MTN not good at all -1

4

MTN, Like seriously, u r the most useless and disgraceful network I've ever experienced in my life... I

regret the day I bought ur chip. I'm porting straight to vodafone and I

‐1

5

so..when..would..u..stop..stealing..my..airtime.? -1

Table 6: Some selected Tigo negative comments.

 Commen

Score

1

tigo is the most useless network in Ghana. bought the ghc39.99 bundle its too slow. I can't even open

common google. you have just lost a customer and I will never

‐1

2

In fact ur internet network is the poorest in Ghana, what network be this?. What it pains me koraa

is....I persuaded my galfriend to port her favorite MTN to Tigo, further on to

‐1

3

masa if u think we re goin to encounter netwrk problems den dont let us activate internet bundles and

later on we cnt use dem.#Poorservice "#Frown uv got tigo"

‐1

4

Tigo Ghana your internet service is really killing me guys, I have a deadline today and I can't even

get anything done. Your Internet is soo slow I can't even open a single page.

‐1

5

hw3 all u pple do is brag. ur service is v.slow. on and off lyk ecg. 0278686875. i will b leaving soon if

u guys dnt gt serious

‐1

Table 7: Some selected Vodafone negative comments.



Comment Score

1

too bad network.. my interent isnt workn again! bomb! im even usin difrnt ntwrk -1

2

stupid network Vodafone for how long will it take your stupid company to fixed my problem my

broadband problem . what the f**k is wrong you with you guys you sick, see

-1

3

And ur internet service has been down for 3 months now.. unbelievable -1

4

a disappointing network, in Takoradi but yet barely able to connect. would not recommend -1

5

And as a company what do you do?when there is a problem.6days of no browsing and nothing has

been.Tweaaaaaaaaaaaaaa what an incompetent data provider

-0.4883

Table 8: Some selected MTN positive comments.



Comment Score

1

Y'ello Emmanuel, As per our discussion,kindly visit any MTN office with an ID and relevant

documents to prove ownership of the number. Thank you.

0.5421

2

Y'ello Vida, Sorry for your experience. Kindly provide your number to enable us investigate your

complaint. Thank you

3

Y'ello Bridget, Thank you for your suggestion,Kindly be informed that it is well noted. Thank you 1

Table 9: Some selected Tigo positive comments.

 Commen

Score

1

We sincerely apologize for any inconvenience caused.

2

Kindly confirm your number and current location for assistance.

0.56801

3

Kindly elaborate on your issue to enable us assist you.

0.68644

Table 10: Some selected Vodafone positive comments.



Comment Score

1

Dial 255 for expert medical advice between 4:00pm and 10:00pm daily. 0.38518

2

In the spirit of love, telecom operator, Vodafone Ghana, has settled the medical bills of over 180

patients who were struggling to raise funds to pay their medical bills through

0.35094

3

Music doesn't lie. If there is something to be changed in this world, then it can only happen through

music.

0.30982

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

addresses specific customer concerns and directly

mentions the name of each customer in the response.

It was also realized in one response that MTN takes

on suggestions of customers and acknowledges the

customer for the input. Tigo also addresses

customers per their concerns. Scanning through the

content, there was some deficiency in Vodafone’s

responses to customer specific queries/enquiries.

The company provided Omnibus statements that

were meant to address general customer issues. To

gain competitive advantage, companies must be seen

to be more responsive to customers concerns.

Companies who personalize customers concerns turn

to retain such customers and increase its competitive

urge over rivals. It can be inferred that MTN has an

urge in terms of responses to customer concerns per

extracted comments on Facebook under the period

of study.

4.2 Sentiment Analysis on Twitter

Adopting the same sentiment analysis score of -1 to

+1 representing most negative to most positive

comments respectively, content analysis of the

actual individual comments and their relative

sentiment score on twitter for the three companies

for randomly selected customers are shown in tables

11-13. Negative comments on twitter for the three

companies are similar to that of Facebook. The

choice of words indicates that these customers are

disgruntled and desire specific response to their

concerns. Content analysis of the positive responses

on twitter was similar to that of Facebook. However,

there were more omnibus statements from the three

companies on twitter as compared to Facebook.

5 LIMITATION

Since the word list in SentiWordNet is not

comprehensive and does not contain vernacular

words some sentiments would not be classified in

the analysis. This affects the accuracy of the

classification. We will therefore develop a wordlist

in future to embody local Ghanaian vernacular

words for a comprehensive analysis.

6 CONCLUSIONS

The explosion of data on social media leaves no

business enterprise with a focus on customer

satisfaction and expansion to be complacent. The

emergence of social media presents businesses with

a rare chance of analysing publicly available data of

their customers/fans and even more importantly that

of their competitors for business advantages. Such

data driven analysis could help businesses to identify

their weaknesses and exploit the weaknesses of their

competitors as far as customer retention and brand

management is concerned It would also help to.

Table 11: Some selected MTN negative comments.



Comment Score

1

Some stupid networks we have in our country @MTNGhana fire burn u all -1

2

u dont care about the plight of Ghanians at all...and also why does it keep sooo long for attendants to

pic calls of customers?

-1

3

u people y?? U pple dey take ma credit small small. I haven't subscribed for any service aside internet

bundles wai. Will port oo

-0.8334

Table 12: Some selected Tigo negative comments.

 Commen

Score

1

Everyday for the last 2months “@Wiredu_: Thou sucketh today @TigoGhana”” << it's becoming

too much. Fix it tiGo

‐0.4674

2

massa, ur internet service dey bore.wk on it den shon dey talk trash.

‐0.8820

3

bt i tot u said 24hrs unlimited. rather 14hrs, such a shame



Table 13: Some selected Vodafone negative comments.



Comment Score

1

too bad network.. my interent isnt workn again! bomb! im even usin difrnt ntwrk -1

2

stupid network Vodafone for how long will it take your stupid company to fixed my problem my

broadband problem . what the fuck is wrong you with you guys you sick, see

-1

3

And ur internet service has been down for 3 months now.. unbelievable -0.7765

EnterpriseCompetitiveAnalysisandConsumerSentimentsonSocialMedia-InsightsfromTelecommunication

Companies

consolidate strengths and identify new opportunities

and threats in the competitive market. In Ghana and

Africa as a whole, social media competitive analysis

may be a new thing but as our research shows, the

numbers of social media users in Africa keep

growing at exponential rates (Deloitte and Touche,

2012; United Nations (UN), 2010).

This study demonstrated how businesses could

take advantage of social media data to improve upon

customer relationship and maintain a competitive

urge over rivals. Using a case study of the top three

Ghanaian Telecom companies, we demonstrated

with sentiment analysis, the general feeling of

telecom users in Ghana. The use of sentiment

analysis is one of the potent ways in attempting to

understand customer behaviour.

The results show how active the three telecom

companies are on social media and how they engage

with their customers/fans on daily basis. Content

analysis of the textual information we extracted

reveal a number insights about how customers

perceive the services of the three companies. Some

of the key patterns and trends we identified are the

following:

(1) Response to User Comments

The study revealed that to stay in touch with their

customers, the three Telecom companies have

specific staff assigned to addressing daily customer

complaints. They use the social media platform to

launch new products and listen to their feedback on

services and products. In responding to customer

concerns, two of the companies were seen to be

giving omnibus answers to complaints. MTN was

innovative in the way they addressed customers by

name and proffered real-time solutions to concerns.

(2) Unusual Posts and Tweets

It was also realized in the content analysis that not

every posts or tweets by the company was related to

their service, products or meant to address a

customer complaint. Some of the companies, TiGo

in particular made several casual postings about

topical issues on their pages to engage their fans in

the day. Most of the topics centered on relationships,

security tips and upcoming entertainment events in

the cities and towns in the country. Though most of

these casual posts yielded lots of responses, we

could not determine their impact on brand awareness

and therefore the competitive urge it gives to such

companies over their competitors.

(3) Service Quality

Though some of the posts and comments from

customers were not business related, majority (60%)

of them were about criticizing or hailing one

company or the other about their service or product

as shown in Tables 5-13. With mobile number

portability service in place in Ghana where users can

port their telephone numbers from one company to

the other for free, competition has become keen

among the various operators about maintaining their

core customer base and capitalizing on the number

portability service to snatch users from their

competitors. In spite of this, some service

complaints were not satisfactorily addressed

prompting others to threaten to port their numbers to

other companies as shown for example by one

comment in Table 11 row 3.

(4) Freebies

We also realized that almost all the Telecom

companies engage in giving out free phone credits to

their customers on social media. Some of these

freebies came in the form of answering simple

questions about their products and services. Some

also came in the form of purchasing tickets to major

entertainments programs for customers who got

some questions rightly answered.

In all MTN seemed to have a slight competitive

urge over their rivals in terms of how the company

engages with customers on social media. A

company’s relative responsiveness to customer

concerns such as MTN makes consumers advocate

of their brand and therefore increase brand referrals

culminating in increased profits and market

visibility.

ACKNOWLEDGEMENTS

This work was supported by Internal Grant Agency of

Tomas Bata University IGA/FAI/2014/037,

IGA/FAI/2014/007 and by the European Regional

Development Fund under the project CEBIA-Tech No.

CZ.1.05/2.1.00/03.0089.

REFERENCES

Aggarwal, R., Gopal, R., Sankaranarayanan, R., & Singh,

P. V., 2012. Blog, blogger, and the firm: Can negative

employee posts lead to positive outcomes?

Information Systems Research, 23(2), 306–322.

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications

Akehurst, G., 2009. User generated content: The use of

blogs for tourism organizations and tourism

consumers. Service Business, 3(1), 51–61.

Bagheri, A., Saraee, M., De Jong, F., 2013. Care more

about customers: Unsupervised domain-independent

aspect detection for sentiment analysis of customer

reviews. Knowledge-Based Systems, 52, 201-213.

Barker, P., 2008. How social media is transforming

employee communications at Sun Microsystems.

Global Business and Organizational Excellence,

27(4), 6–14.

Barlow, M., Thomas, D. B., 2010. The executive's guide to

enterprise social media strategy: how social networks

are radically transforming your business (Vol. 42).

John Wiley & Sons.

Bradley, A. J., McDonald, M. P., 2011. The Social

Organization: How to Use Social Media to Tap the

Collective Genius of Your Customers and Employees.

Harvard Business Press.

Chavula, H. K. 2013. Telecommunications development

and economic growth in Africa. Information

Technology for Development, 19(1), 5-23.

Dave, K., Lawrence, S., Pennock, D. M., 2003. Mining the

peanut gallery: Opinion extraction and semantic

classification of product reviews. In Proceedings of

the 12th international conference on World Wide Web

(pp. 519-528). ACM.

Deloitte, Touche, 2012. Social Media Africa. Africa

Frontiers Forum

Dyer, P., 2013. Top Tools for Social Media Monitoring,

Analytics, and Management. Retrieved from:

http://socialmediatoday.com/pamdyer/1458746/50-

top-tools-social-media-monitoring-analytics-and-

management-2013

Esuli, A., Sebastiani, F., 2006. SentiWordNet: A publicly

available lexical resource for opinion mining. In

Proceedings of LREC (Vol. 6, pp. 417-422).

Essoungou, A.M., 2010. A social media boom begins in

Africa. United Nations, Africa Renewal. Retrieved

from:http://www.un.org/africarenewal/magazine/dece

mber-2010/social-media-boom-begins-africa

Feldman, R., 2013. Techniques and applications for

sentiment analysis. Communications of the ACM,

56(4), 82-89.

Ghiassi, M., Skinner, J., Zimbra, D., 2013. Twitter brand

sentiment analysis: A hybrid system using n-gram

analysis and dynamic artificial neural network. Expert

Systems with Applications, 40(16), 6266-6282.

Gillin, P., Schwartzman, E., 2011. Social marketing to the

business customer: Listen to your B2B market,

generate major account leads, and build client

relationships, John Wiley & Sons. New Jersey, 1

edition.

Governatori, G., Iannella, R., 2011. A modeling and

reasoning framework for social networks policies.

Enterprise Information Systems, 5(1), 145–167.

Harvard Business Review, 2010. The New Conversation:

Taking Social Media from Talk to Action. SAS

Institute. Harvard.

Jussila, J. J., Kärkkäinen, H., Aramo-Immonen, H., 2014.

Social media utilization in business-to-business

relationships of technology industry firms. Computers

in Human Behavior, 30, 606-613.

Kang, D., Park, Y., 2014. Review-based measurement of

customer satisfaction in mobile service: Sentiment

analysis and VIKOR approach. Expert Systems with

Applications, 41(4), 1041-1050.

Koppel, M., Schler, J., 2006. The importance of neutral

examples for learning sentiment. Computational

Intelligence, 22(2), 100-109.

Lipsman, A., Mudd, G., Rich, M., Bruich, S., 2012. The

power of ‘‘like”: How brands reach (and influence)

fans through social-media marketing. Journal of

Advertising Research, 52 (1), 40–52

Mahmoud, M. A., Hinson, R. E., 2012. Market orientation,

innovation and corporate social responsibility

practices in Ghana's telecommunication sector. Social

Responsibility Journal, 8(3), 327-346.

Mittal, B., Lassar, W. M., 1996. The role of

personalization in service encounters. Journal of

retailing, 72(1), 95-109.

NCA-Ghana, 2013. Review of mobile and fixed network

services statistics for the year 2013. Retrieved from:

http://www.nca.org.gh/73/34/News.html?item=357

Pang, B., Lee, L., 2008. Opinion mining and sentiment

analysis. Foundations and trends in information

retrieval, 2(1-2), 1-135.

Pang, B., Lee, L., Vaithyanathan, S., 2002. Thumbs up?:

sentiment classification using machine learning

techniques. In Proceedings of the ACL-02 conference

on Empirical methods in natural language processing-

Volume 10 (pp. 79-86). Association for Computational

Linguistics.

Peppers, D., Rogers, M., 1995. A new marketing

paradigm: share of customer, not market share.

Managing Service Quality, 5(3), 48-51.

Qualman, E., 2009. Socialnomics: how social media

transforms the way we live and do business. Wiley

John & Sons, Inc. Hoboken, 2

edition.

Safko, L., 2010. The Social media bible: tactics, tools, and

strategies for business success. John Wiley & Sons.

Hoboken, New Jersey, 2

edition.

Sashi, C. M., 2011. Customer engagement, buyer-seller

relationships and social media. Management Decision,

50(2).

Sinderen, M. V., Almeida, J. P. A., 2011. Empowering

enterprises through nextgeneration enterprise

computing. Enterprise Information Systems, 5(1), 1–8.

Taboada, M., Brooke, J., Tofiloski, M., Voll, K., Stede,

M., 2011. Lexicon-based methods for sentiment

analysis. Computational linguistics, 37(2), 267-307.

Thelwall, M., Buckley, K., Paltoglou, G., 2012. Sentiment

strength detection for the social web. Journal of the

American Society for Information Science and

Technology, 63(1), 163-173.

Tong, R. M., 2001. An operational system for detecting

and tracking opinions in on-line discussion. In

Working Notes of the ACM SIGIR 2001 Workshop on

Operational Text Classification, Vol. 1, pp. 6.

EnterpriseCompetitiveAnalysisandConsumerSentimentsonSocialMedia-InsightsfromTelecommunication

Companies

Turney, P. D., 2002. Thumbs up or thumbs down?:

semantic orientation applied to unsupervised

classification of reviews. In Proceedings of the 40th

annual meeting on association for computational

linguistics (pp. 417-424).

Weber, L., 2009. Marketing to the social web: How digital

customer communities build your business, John

Wiley & Sons. Hoboken, NJ, 2

edition.

Wiebe, J. M., 1994. Tracking point of view in narrative.

Computational Linguistics, 20(2), 233-287.

Wollan, R., Smith, N., Zhou, C., 2010. The social media

management handbook: everything you need to know

to get social media working in your business. John

Wiley & Sons. Hoboken, 1

edition.

Zeng, L., Li, L., Duan, L., 2012. Business intelligence in

enterprise computing environment. Information

Technology & Management, 13(4), 297–310.

Zhang, W., Xu, H., Wan, W., 2012. Weakness Finder:

Find product weakness from Chinese reviews by using

aspects based sentiment analysis. Expert Systems with

Applications, 39(11), 10283-10291.

DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications