Development of a Conceptual Framework for Machine Learning Applications in Brick-and-Mortar Stores
WI2020 Zentrale Tracks
The growing prevalence and impact of e-commerce puts traditional brick-and-mortar stores under pressure. More and more customers prefer the variety of goods, easily comparable prices, and personalized recommendations online to conventional shopping experiences in stationary retail. A major asset of online stores is their potential to collect, analyze, and interpret data. The collection and analysis of customer behavior and transaction data to improve website design, the assortment, and pricing
... tment, and pricing strategies so-called 'web analytics' are common practice in e-commerce for more than fifteen years already. Advancements in technologies and the ongoing digitalization of brickand-mortar stores unveil the potential of Retail Analytics for conventional stores as well. Yet, a structured overview of diverse factors relevant for implementing Retail Analytics is missing. In light of this context, this article derives a conceptual framework harmonizing the relations between different technologies, collected data, analysis methods, method outputs, and application purposes. stores usually operate on a single physical channel without adopting new technologies within their stores. A major asset of online stores, leveraging their dominant position, is their immense potential to collect, analyze, and interpret data about their businesses. The collection and analysis of customer behavior and transaction data to improve website design, the product assortment, and pricing strategies so called 'web analytics' are common practice in e-commerce for more than fifteen years already  . These techniques have considerably contributed to the success of e-commerce. Davenport's Harvard Business Review article "Competing on analytics" , which has been cited more than a thousand times, underlined the immense power of data analytics for business purposes and can be seen as the first seminal paper on this topic leading to the emergence of a whole new research area, which has significant relevance for stationary retail as well. Yet, most brick-and-mortar stores restrict their analyses to simple methods like implementing customer counters or sending test customers into their stores, while further customer touchpoint within the store are not investigated . Furthermore, brick-and-mortar stores are not necessarily used to data-driven decision making, which would lead to better decision making. Since the domain knowledge as well as experts, applying their knowledge, are missing, they are in need for support, which could offer additional benefits to them and their customers. During the literature search, nearly no real holistic, structured overview over possible opportunities for brick-and-mortar stores to adopt new technologies or machine learning could be found. However, multiple sources show single implementations of technologies and/or machine learning in the context of brick-and-mortar stores, providing them with examples to improve their business. An overview over these applications could offer brick-and-mortar stores the first point of contact in order to engage with different ideas and possibilities. This led to the question of how brick-and-mortar stores could be supported in discovering and adopting new technologies as well as machine learning in order to offer new customer experiences as well as smart, data-driven decision making? To solve this question, this article aims to offer a structured overview of existing applications. Thus, showing brick-and-mortar stores what they can achieve by implementing machine learning with their currently present data as well as possible technologies they could introduce to improve their business. This can guide brickand-mortar stores on their way to discover new technologies and implement machine learning. Additionally, it also closes a research gap by providing a structured overview over the topic of applied machine learning in the context of brick-andmortar stores. The remainder is structured as follows: section 2 delineates our research design and section 3 presents the research background. In section 4, the Retail Analytics framework for brick-and-mortar stores is explained. The discussion and evaluation are presented in section 5 followed by limitations, implications for research and practice and the conclusion in section 6.