Generating Value Using Predictive Analysis in E-retail : A Case Study on How Predictive Analysis Affords Value-Generating Actions

Abstract: Information systems and information technology are rapidly evolving, and the usage of it at the same pace. In different fields, predictive analysis is used daily. Within the area of e-retail, referring to online retailing, it is used for personalisation and as decision support. There’s a lot of research on how to increase the accuracy of the predictions and different methods for this, however, there’s a lack of research regarding the actions an organisation can take given different predictions. Hence, this master’s study researches what factors affording or constraining value are in relation to the usage of predictive tools given different organisational roles. This thesis is made by a case study with a qualitative approach, following the interpretivism paradigm. The data used in this research comes from document analysis followed by semi-structured interviews to gather additional information about what the document analysis or previous research has not covered. The empirical findings were analysed using thematic analysis and are then discussed in relation to the research questions and theoretical framework, together with what’s previously been stated in the literature. The research questions for this thesis are the following: RQ1: How do different organisational roles affect the actions taken on information from predictive analysis in e-retail? and RQ2: What are the key factors affording or constraining value generation in predictive analysis within e-retail? The empirical findings resulted in six themes, where three are relevant to each research question. The findings suggests that there are four major categories of roles that have similar affordances of predictive analysis, these are customer-, sales and financial operations-, management-, andlastly supply chain and inventory related. When several roles within an organisation use the same prediction tool, there are positive effects such as less biased decisions and improved communication through collaboration. Several factors, both constraining and affording value were found. The main constraining factors are related to technological knowledge and interpreted value as well as trustworthiness. The affording factors are instead the allowance of tying predictions to certain KPIs and the ability to be able to slice into the data to show what’s relevant for the individual. In addition to these factors, some desires for functionality were found. These were, among others, a confidence score of the predictions, prediction for certain goals, and predicted optimal send times for emails in the future. My suggestion for future research is to approach the same problem using another theoretical framework to further enhance a novel field, as well as involving participants with different backgrounds than was used in this thesis.