Direct-to-consumer eCommerce data strategy and sales forecasting with neural networks

Abstract

The emerging business model of direct-to-consumer (D2C) has disrupted and transformed the eCommerce industry. An increasing number of companies are handling the design, manufacturing, marketing, sales, and shipping of their products themselves. Success in the D2C eCommerce business model relies on a thorough understanding of consumer behavior and the ability to effectively target marketing efforts to the appropriate audience. Therefore, D2C eCommerce companies need to have a solid eCommerce data strategy defining what metrics to gather, what metrics to use to steer business, and how to collect and combine relevant data. This master’s thesis consists of blueprinting D2C eCommerce data strategy elements and examines eCommerce sales forecasting using neural networks. The objective of this thesis is to preform literature review for D2C eCommerce companies' data strategy elements including baseline key metrics, appropriate data sources, and artificial intelligence/machine learning methods for initiating data-driven D2C eCommerce. Additionally, the thesis investigates the feasibility of neural networks in predicting eCommerce sales. The literature review consists of eCommerce data strategy elements, followed by the blueprint of eCommerce data solution architecture, introduction of various statistical methods usable for time series data analysis, and neural networks theory. The sales forecasting was performed using neural networks, and the findings were assessed using evaluation metrics. Sales forecasting research results outlined the difficulties of predicting sales accurately with neural networks in turbulent eCommerce business. Although advertisement cost and sales had a correlation, a relatively small dataset and highly fluctuating sales made sales forecasts inaccurate. However, the literature review of research emphasised the value of a robust D2C eCommerce data strategy in enhancing business performance.