The application of business analytics techniques to analyze unstructured text from various sources to complement state-of-the-art opinion leaderidentification and management in the european public procurement law

Resum

The objective of the dissertation is to analyze tenders with the techniques of business analytics in order to uncover characteristics narrowing the competition in the specification for tenders resp. documents in which the requirement for competition was bypassed. Consequently, this has an impact on the probability of a provider being awarded a contract and at the same time discloses the motivation for actively influencing a tender. Among others, this verifies the assumption that a large part of public tenders has already been influenced by opinion leaders before they are published. It is not only possible to provide evidence of this influence with business analytics techniques, the assertion of influence also enables the identification of the opinion leaders involved. Modern business analytics procedures provide for an extensive analysis of the tender documents and at the same time a prediction of the probability of being awarded a contract. Therefore, the main research question referred to the findings on opinion leaders to be acquired regarding their significance and influence in the context of the public tender using business analytics techniques. Public tenders are of particular interest, as the public sector is one of the biggest demanders on the German market with an annual volume of about 496 billion euro. The IT sector alone scheduled expenses in the amount of 20.9 billion euro for 2014. For this reason the economic interest driving the influencing of a tender should not be underestimated. Parts of the examination have already been published in papers by the author of this dissertation. The line of argument is based on a segment of capital goods that is associated with significant cognitive control and characterized by an influencing approach that differs from the patterns of the mass market. Within these segments the identification was focused on the complex capital goods of information technology. This segment is shaped by a number of characteristic features, including great innovational strength, short product lifecycles, disruptive technologies, low market entry barriers and insufficient market transparency, among others. These characteristics make it possible for experts with domain-specific know-how to establish an inverse connection between the product-specific features and the number of possible vendor addressees. As already pointed out, the opinion leader assumes a key role in the entire examination, as his higher level of information places him in the center of the process of influencing of tender. His relay and multiplier function as well as extensive knowledge enable him to assert cognitive and emotional influence and to assume a key position in the process of discussion and evaluation. Especially the topic-specific multiplier function for product and brand messages is of particular interest for companies. Another role of the opinion leader, which is significant in particular for this dissertation, is the influencing function, e.g. to eliminate behavioral and decision-making insecurities. This is to be verified using the business analytics procedures. Within the framework of the examination a higher-level process model was developed to answer the research question experimentally applied to the “storage” system. The purpose was to achieve five derived objectives by applying business analytics procedures. The first objective was to identify opinion leaders. This was successfully achieved by means of a correlation analysis using the text mining approach on different sources of information. The second and third goals involved extracting vendor-specific functionalities that lead to a narrowing of the competition and therefore an evaluation of the level of influence asserted on a tender. Using an especially developed domain-specific taxonomy based on 697 extracted indicators, it was possible to transfer 495 tenders to a model for displaying the degree of influencing. 86% of all analyzed documents showed a significant narrowing of the competition or clearly vendor-specific characteristics. Only 14% of the tenders were actually neutral resp. only displayed a minor narrowing of the competition. The fourth objective involved the prediction of the probability for each tender that a vendor is rewarded the contract. The tools of regression, the decision tree and the neural network were applied in the course. The most reliable predictions based on the misclassification rate were achieved by the neural network and by the regression approach. 15 expert interviews were conducted with vendors, partners and customers in order to triangulate and validate the findings from the applied business analytics procedures. Subsequently, the more than nine hours of audio material were interpreted using the method of qualitative content analysis. The interview focused on the content design of a tender as well as the actual process, the information search as well as questions regarding the experts’ self-designation. Thanks to the correlation of the entire material, it was possible to create an all-encompassing reconstruction of the procurement process, while shedding light on different strategies of influencing a tender. An early limitation of the number of bidders, e.g. by providing a description of an operational concept, and creating artificial USPs are the most prominent examples of active influencing. Furthermore, it was possible to design a sociometric network that displays the influencing communication channels and protagonists in detail. By triangulating the findings from the expert interviews it was possible to confirm the detected patterns in the tender documents using the procedure of business analytics. Moreover, the interviews revealed additional influencing strategies that could not be evaluated in the context of text mining, but which should be considered as part of the influencing. As a conclusion, in consideration of all aspects referred to in this dissertation, it is possible to identify influence asserted by vendors in many public tenders, in addition to the fact that this exercise of influence can be traced back to the opinion leader with the techniques of business analytics. In any case the identification of the opinion leader, or knowing about the act of influencing, offers a number of interesting consequential courses of action. These include opinion leader management and the purposeful application of resources resp. an adjusted pricing strategy when replying to a tender.