In recent years, there has been an outstanding and overwhelming growth in technology and its application within organisations. Many of these organisations have adopted technology not only to have easier execution of processes, but  to  be  at  a  better  advantage  than  their competition. Linking competitive advantage to business intelligence (BI) has emerged as one of the key topics in organizational science and information systems (Chen et al., 2012).

BI has seen many enhancements and significant improvements in the past few years. Strategically and from a decision making perspective, BI assists organizations in corporate performance management; optimizing customer relations, monitoring business activity, and traditional decision support; packaging standalone BI applications for specific operations or strategies and management reporting of business intelligence (Negash, 2004). Organisations use BI as a tool for these processes to produce measurable and quantifiable factual data to support their strategic decisions. To plan their strategies, these organisations need to consider the pressures and challenges caused by the business environment in order to succeed and thrive in their industry and the global economy as a whole (Pirttimäki and Lönnqvist, 2006).

In the current economic climate, organisations strive to find ways to be better than the competition and to remain profitable. A business intelligence approach provides organisations with the ability to provide information that assists in strategic planning and decision making, and is based on data specific to the organisation. This study is focused on exploring the following research question: How can organisations use sentiments analysis in business intelligence and knowledge management to build a sustainable competitive advantage? The main objective of this paper is to investigate the use of sentiment analysis to gain competitive advantage. The relationships between the constructs of business intelligence, knowledge management, and sentiment analysis are investigated. We also investigate how organizational structure influences or compliments the relationships between these constructs.

The section also looks at the concept of sentiment analysis as a separate concept and how it integrates into both business intelligence and knowledge management. Competitive advantage is also explored and how it is linked to business intelligence, knowledge management, and sentiment analysis.

Business intelligence and knowledge management

Luhn (1958) describes intelligence as the ability to comprehend the interrelationships of presented facts in such a way as to guide action towards a desired goal. Over the years, BI has consequently borrowed from Luhn’s description of intelligence and has been described as the ability by organisations to take all their historic data (and current data) and convert them into useful information. Ultimately, this information, through analysis, is converted into knowledge which is then conveyed to the organisations’ main stakeholders and decision makers  as  promptly  as   possible.   The   question   has always been, “how can we use the information at our disposal better to assist in making better informed decisions?” BI primarily helps provide historical, current and predictive views of business operations from data stored in the data warehouse, which provides organizations with an opportunity to make effective business decisions(Wang and Wang, 2008). These technologies, processes, and applications analyze structured and unstructured organization data and business processes within the organisation. Demand for BI applications continues to grow even at a time when demand for most IT products is soft (Parenteau et al., 2016). According to Wang and Wang (2008), organisations utilising massive data to gain competitive advantage seems to be the central theme of business intelligence deployment.

There is a perception that business intelligence and knowledge management are two independent information systems domains (Wang and Wang, 2008).  Elbashir et al. (2008) make the case that with business intelligence, organisations can improve decision making and through better informed decisions, gain competitive advantage. While this is true, Herschel (2005) argue that knowledge management can also be used to support decision making processes by providing new ways for sharing knowledge in organisations. Pollak et al. (2012) argue that these two fields may offer considerably more and better benefits if they are in synergy; with their integration resulting in improvement in decision support capabilities in organisations and promote the formation of ‘organisational memory’ repository. Wang and Wang (2008) put emphasis on the potential of integrating these two fields. Early efforts to find synergy between the two fields also considered data mining, not only as a business intelligence tool, but that the process of data mining is also a knowledge management process (Chen and Liu, 2005). The elaboration of data mining as a tool and a process results in a natural connection between business intelligence and knowledge management (Wang and Wang, 2008).

Sentiment analysis and opinion mining

Sentiment analysis or opinion mining is the computational study of natural language (including opinions, sentiments, and emotions) by applying computational linguistics, and text analytics to identify and extract subjective information from opinions by human beings (Mejova, 2009).

The objective is to try to understand opinions, whether spoken or written, with the aid of technological applications. Like business intelligence, it is used to gain insight, generally from customer and consumer commentary and opinions in order to help organisations using it. The techniques and technologies for sentiment analysis have been adapted in such a way that allows them to be automated allowing  optimum,  enhanced  and advanced sentiment analysis through automated sentiment analysis. There are different natural language approaches that allows automated process to detect the sentiment of short informal textual messages such as tweets and SMS (message-level task); and the sentiment of a word or a phrase within a message (term-level task) (Kiritchenko et al., 2014). There are also several measurement platforms which employ different statistical methodologies and techniques to evaluate sentiment which may be found across the web in various forms (You et al., 2016). A key challenge in sentiment analysis as a classification process is determining whether a particular sentiment is subjective or not, and whether the opinion expressed is positive or not (Pang and Lee, 2004; Medhat et al., 2014).

Classification is a key challenge since sentiments and opinions are predominantly unstructured in form, complex and large, existing in various databases, social media sites and in corporate websites. In linking sentiment analysis and opinion mining, we recognize that the competitive environment of organizations has changed and there is need to monitor and analyze not only the customer-generated content on their own social media sites, but also the textual information on their competitors’ social media sites (He et al., 2013). Thus, sentiment analysis and opinion mining is increasingly forming part of the repertoire of tools that are used for generating competitive advantage.

Linking sentiment analysis to business intelligence and sustainable competitive advantage

Competitive advantage is implementing a strategy currently not used by the competition, whereas sustainable competitive advantage implies that the implemented strategy cannot be duplicated (Barney, 1991). Duplication of strategy is made much more difficult if resources rare, non-substitutable, valuable, and inimitable. Organisations must develop and nurture core competencies in order to establish and sustain their competitive advantage (Prahalad et al., 1990). Organisations need to invest in capabilities such as active information acquisition, incorporation of customer’s voice/opinions, and knowledge sharing and distribution (Kumar et al., 2011).

We envisage the ‘glue’ to linking sentiment analysis to sustainable competitive advantage to the knowledge management perspective in which knowledge is seen as a strategic asset with the potential to be a source of competitive advantage for an organization (Halawi et al., 2005). Thus it is necessary for organizations to innovate from the generic thinking of competitive strategy, and instead consider the use of knowledge management. Additionally, knowledge management enables the organisations to remain competitive in the changing environment with more focus on adaptability, survival, and abilities (Rahimli, 2012). Thus the key  to  developing a sustainable competitive advantage lies in the ability to acquire and to transfer knowledge, which, as we had seen earlier, can be realized through the process of data mining.

Fuloria (2011) discusses the use of advanced analytics by organisations to maximize profits and build long-term sustainable competitive advantage. He identifies four forces that drive advanced analytics: consumer deleveraging, persistent consumer frugality, increased regulations and the surfeit of data. There is currently a surfeit of data generated by users connected to the World Wide Web (WWW) in various different platforms like social networks, blogs, forums, and products websites through different types of devices like mobile phones and desktop computers. Most of the data generated in these platforms is in a form of sentiment. This torrent of data has led to the evolution of analytics from just reporting to predictive analytics (Fuloria, 2011). As it has been discussed in the literature above, knowledge is a key component to building a sustainable competitive advantage. Business intelligence has also been identified as a key component to a sustainable competitive advantage. Integrating both business intelligence and knowledge management would enable organisations to create niche business intelligence in order to maintain a competitive advantage. We therefore seek respondents’ opinions on the integration of the constructs of Sentiment Analysis, Business Intelligence, Knowledge Management and Sustainable Competitive advantage in this study.

The research conducted was limited to South African Companies. Furthermore, it was limited to the knowledge of IT project managers and knowledge management experts working in projects geared towards realizing the use of BI in organizations. The researchers used a quantitative survey research strategy. Quantitative research is primarily used by positivist researchers and a survey is strongly linked to positivism. The sampling frame comprised of 79 organizations out of which responses were received from 30 firms from various sectors of South Africa. A convenience sampling technique was employed to identify those organizations that are currently using BI or some knowledge management application. The representatives of these companies comprised of a cohort of honors and masters part time students, employed in those organizations and knowledgeable about the focus of the study. The demographic profile of the respondents ranged from graduate trainees to senior managers in the 30 organizations that participated in the research.

Key constructs

The key constructs, motivated in the literature review section, were Business Intelligence, Knowledge Management, Sentiment Analysis/Opinion Mining and Competitive Advantage. Business Intelligence, Knowledge Management and Sentiment Analysis items were considered as the independent variables; while the Competitive Advantage items were the dependent variables (Table 1).  Although the test items were sourced from instruments used in prior research (as per Table 1), the instrument was piloted among 5 academics and minor corrections for purposes of clarification and better language use were done.

Data analysis tools

A descriptive and inferential analysis was used to identify patterns in the data and draw conclusions. A factor analysis based on a principal axis factoring with varimax rotation of the scale was conducted to investigate the internal structure as well as to determine the smallest number of factors that could be used to best represent the interrelations among the sets of variables for the construct. In deciding on the number of factors to extract, a combination of the Kaiser-Guttmann Rule (K1 rule), and the scree plot were utilized to determine the most appropriate component solution. 

Reliability analysis

In order to assess whether the data from which the 7 variables of Business Intelligence were reliable, Cronbach Alpha’s were computed. The alpha for the 7 items of BI was 0.927, and was therefore retained for further analysis. The Knowledge Management (KM) construct had a Cronbach Alpha of 0.947, but had a number of the items scoring below 0.6, implying that their inclusion into the set was questionable. These were KM8 (0.643), KM 9 (0.668), KM10 (0.595) and KM13 (0.698). These four items were therefore deleted and were not considered for further analysis. The Opinion Mining (Sentiment Analysis) construct had 11 items, with a Cronbach’s Alpha of 0.924; but with 7 items having coefficients between 0.7 and 0.6, which are therefore considered questionable. These seven items (OM1=0.683; OM4=0.605; OM5=0.648; OM6=0.638; OM7=0.683; OM10=0.637 and OM11=0.574) were deleted and were not considered during further analysis. 7 of the 8 the Competitive Advantage variables had Cronbach Alphas of over 0.7 and were therefore acceptable, while the overall coefficient for the construct was 0.942. Thus after assessing the internal consistency of the test items and deleting those items whose coefficients fell below a certain threshold, the total number of variables that remained was 27 (Table 2). 

The new 27 – item scale was factor analyzed and the resulting optimal factor solution interpreted. A summated scale was then constructed to form the basis for subsequent multiple regression analysis. According to Tabachnick and Fidell (2001) factors with a single variable can be described as poorly defined. Factors with two variables should be highly correlated with each at weightings of > .70. The optimal solution contained four factors, with an explained variance of 74% (Table 3). 

In order to test for the suitability of the data to factor analysis, the Kaiser-Meyer-Olkin measure of sampling adequacy and Bartlett’s test were used. The Kaiser-Meyer-Olkin Measure of Sampling Adequacy is a statistic that indicates the proportion of variance in the variables that might be caused by underlying factors. The closer the value of KMO to 1.0, the more useful the results of the factor analysis are. Table 4 indicates that KMO value is 0.47. We indicated though that this was an exploratory factor analysis and the intention was to use it as a heuristic tool for uncovering the underlying structure. Bartlett’s test shows a value of 0.000, which is less than 0.05, indicating that the data is amenable for factor analysis (Table 4).

Table 5 presents the items, factor loadings, descriptive statistics and the names that were given the variables that loaded on the factors. The naming of the factors took into account the significance of the loadings. Factor 1, highlighting the Knowledge Management component, had six variables (KM7, KM6, KM3, KM2, KM4, and KM12) heavily loaded. The three heavily weighted variables focuses on integration of different types (KM7) and sources (KM6) of knowledge in the development of key organizational strategies (KM3). The other three KM processes seek to use KM in decision making (KM2), acquisition of customer knowledge (KM4) and in improving organizational performance. Given the high loadings (Table 5) recorded for the six variables, the means of the six KM variables will be used as surrogate values for further regression analysis in the next section. 

Factor 4, named Opinion Mining/Sentiment Analysis, comprised of four items (OM2 = 0.87; OM3 = 0.79; OM8 = 0.78; OM9 = 0.77) reiterating the importance of processes for acquiring, filtering, using and matching customer opinions to challenges and problems. Thus the opinion mining variables simply emphasize the need for developing the process of integrating sentiment analysis into organizational processes.  The four factor interpretation above suggests that three summated scales can be constructed to represent the three constructs of KM, BI and OM representing the independent variables; while a single summated scale for the CA construct is applicable to the dependent variable (Table 6). Each of the summated scales uses a surrogate variable with the highest factor loading. The reliability of the summated scale is best measured by Cronbach’s alpha, which in this case is 0.94 for scale 1 (factor 1), 0.96 for scale 2 (factor 2), 0.86 for scale 3 (factor 3) and 0.90 for scale 4 (factor 3). 

Multiple regression analysis

The 3 – item summated scale for KM, OM and BI representing the set of multiple independent variables and the 1 – item summated scale representing the set of multiple dependent variables was used in further multiple regression analysis, as a set, to explain the relationship between the predictor (IV) and the outcome variables (DV). The statistical problem involves identifying the extent and nature of the link between the underlying latent relationships between the set of dependent and independent variables. The analysis in this study provides evidence linking BI, KM and OM to sustainable CA. The correlation matrix of the summated scale variables had significant correlations between the IVs and the DV, but with insignificant correlations between the IVs. The insignificant correlations amongst the IVs confirm that there are no multi-collinearity concerns in the remaining data for further regression analysis.

The model summary (Table 7) shows that the multiple correlation coefficient (R), after deleting insignificant predictors and using KM7 and B12 simultaneously, is 0.715 (R2 = 0.38), and the adjusted R2 is 0.476, meaning that 48% of the variance in Competitive Advantage can be explained and predicted by Knowledge Management and Business Intelligence. According to Cohen, Cohen, West, & Aiken (2013), this is a large effect. This explanation is in part related to the fact that only proxy variables were used as well as the magnitude and the effect of the sample size (29). The proxy variables were used to minimize the effects of collinearity, which was evident when all the variables for each construct used. 

The ANOVA summary (Table 8) shows that F = 14.146 and is significant, which gives an indication that the combination of the predictors significantly explain and predict competitive advantage.  The correlations are shown in Table 9 which shows that the two predictors are significantly correlated to the independent variable. Table 10 suggest that the two predictors (KM and BI) are significantly contributing to the equation (see the sig column); thus implying that knowledge management and business intelligence contributes most to the attainment of competitive advantage, barring other factors unrelated to the constructs that were being taken into account in this study. Table 10, under collinearity statistics, also shows that we do not need to worry about collinearity since the VIF values are close to 1, thus there is  minimal multicollinearity. Therefore, from a predictive perspective, KM and BI are critical for realizing CA; while OM has an insignificant role toplay. 

The results from the regression and correlation analysis linked knowledge management, business intelligence and competitive advantage. The knowledge management factor specifically highlighted that an organization requires processes of integrating different types of knowledge (KM7) and processes that utilizes business intelligence to improve organizational performance (BI2). These two processes, based on the model results are likely to impact on sustainable competitive advantage if an organization has a unique opinion utilization processes that are not easy for the competition to imitate (CA6). The impact of Knowledge Integration Processes, which had a factor loading of 81%, can be interpreted to be linked to the quest by organizations to achieve integration of the knowledge assets by establishing processes to optimize knowledge transfer within organizations. This finding is not new, and various researchers (Lee & Choi, 2003; Nahapiet & Ghoshal, 1998)have elaborated on the significance of knowledge processes as a foundation to organizational performance and one of the key capabilities is developing processes for knowledge transfer. However, the value from the finding links to the need not only to transfer knowledge, but also to consider types (characteristics) of knowledge transferred.

A possible implication is that, while development and integration of the processes of knowledge transfer is vital, the other aspect is to recognize that organizational knowledge assets may be diverse and the nature of these assets need to be understood. Again, there have been studies that have been done documenting the types of knowledge and what we take from the import of the finding is that the processes of knowledge transfer need to take into account the type of knowledge. Findings by Evans & Easterby-Smith (2001)identify how various knowledge types can be embedded in the process of knowledge transfer. So, in a preliminary sense, and  from a developing country’s perspective, we recognize the importance attached to the influence of knowledge integration process on competitive advantage.

The other influence factor related to business intelligence (BI2) had a factor loading of 75% in the factor analysis, was significant (at 0.006) and had the largest beta value in the regression model. The business intelligence factor focused on the need to develop processes that utilize business intelligence to improve organizational performance. As it has been discussed in the literature, knowledge is a key component to building a sustainable competitive advantage. Business intelligence has also been identified as a key component to a sustainable competitive advantage. Cheng, Lu, & Sheu (2009)confirms that integrating both business intelligence and knowledge management would enable organisations to create niche business intelligence in order to maintain a competitive advantage, thus impacting on organizational performance. Business intelligence, considered as a data mining tool, and knowledge management, a process of data mining (Brachman and Anand, 1996), can impact on organizational performance. 

This study was premised on the notion that opinion mining/sentiment analysis, knowledge management and business intelligence plays a role in influencing the competitiveness of organizations. The starting point was to theorize that there are linkages between the independent constructs of sentiment analysis, knowledge management and business intelligence with the dependent construct of sustainable competitive advantage.  Firstly, the components making up sentiment analysis, knowledge management, business intelligence and competitive advantage categories were derived empirically using exploratory factor analysis. Even though the survey instrument was based on validated instruments from prior literature, it was found that the latent factors which were eventually uncovered did not map very well on the factors originally postulated in the literature.  In   particular,   the   knowledge   management cluster of items had four items out of seven having Cronbach alphas of less than 0.60 and were therefore dropped from the item set. In addition, items representing sentiment analysis had set items dropped due to their Cronbach alpha score; while only one item was dropped for those items representing competitive advantage. Thus from an item set of 37 items, 10 items were deleted based on “rules of thumb” related to Cronbach alpha resulting in a final item set of 27, with 30 valid cases used for analysis.

A further factor analysis of the 27 items that remained was undertaken to extract the latent factors and achieve reduction of the variables that can be used to increase the explanatory power. Four factors were confirmed extracted, with items loading correctly to the early theorized constructs. Under the construct of knowledge management, there were six items that were significantly loaded with the emphasis being the integration of knowledge sources for the development of organizational strategies. Under the construct of business intelligence, two variables were significantly loaded emphasizing the need to integrate knowledge management and business intelligence for better decision making to realize organizational performance. A third factor, focusing on sentiment analysis, had four factors highly loaded, orientated towards the need for organizational processes for acquiring, filtering, using and matching customer opinions to challenges and problems. Thus, the opinion mining variables simply emphasize the need for developing the process of integrating sentiment analysis into organizational processes. The fourth factor, consisting of the dependent variables, with four variables heavily loaded. The fourth factor was related to the utilization of opinion, KM and BI processes and an integration of BI and KM for sustaining competitive advantage. Thus as per the factor analysis, all the highly loaded items of competitive advantage captured an aspect of opinion mining, KM, BI and the integration of these processes. Thus, the use of factor analysis preliminarily confirms that the theorized constructs of KM, BI and OM are latent constructs that can be used to explain and possibly to predict sustainable Competitive Advantage in organizations.

Secondly, a regression analysis was undertaken after using a summated scale of the reduced variables from the factor analysis results above. Surrogate variables were used to represent KM, BI, OM and CA constructs. An analysis of the reliability of the summated scale was confirmed, after which proxy variables were used to represent the various constructs. The multiple regression analysis undertaken sought to explain the nature and the strength of the link between the independent and the dependent variables. The regression model summary highlighted earlier showed that the surrogate variables related to KM and BI can be used to explain CA; while OM was insignificant, taking into account the sample, its size   and   the   context   within   which   the   study   was conducted. The finding may appear contradictory to literature, implying that OM may not be contributing to sustainable competitive advantage. However, on closer scrutiny, the surrogate variable that represents the dependent construct of sustainable competitive advantage directly ‘speaks’ to the role of OM orientated towards processes that utilize opinions geared towards integration of KM and BI for sustaining competitive advantage. Thus, while the regression model depicted OM as insignificant in CA, the significant surrogate variable used for representing the DV specifically captured the import of OM. OM, while negated as an independent variable, found its way as the ‘glue’ in linking BI and KM to sustainable CA. So overall, the study confirms the criticality of the dependent variable constructs in influencing the sustainability of organizational competitive advantage; though not as early theorized.

A number of limitations must be recognized. Firstly, the sample size is relatively small, though the exploratory nature of the study allows us to make the preliminary conclusions that we make in this research; but we do recognize that the statistical support for the proposed constructs is not as strong. There is no evidence for the generalizability of the specific findings in terms of factors uncovered to other developing countries where the context may be very different. Secondly, although a theoretical rationale is given for the theoretical model whereby the constructs appear to be linked, it must be recognized that, from a statistical point of view, the support is only for correlation (or association) between the factors as highlighted and the directionality cannot be substantiated. Future research is vital.

The authors have not declared any conflict of interests.