Selecting an Optimal Stream Processing Tool in an E-commerce Environment

Abstract

The rapid growth of data volume and velocity in e-commerce has heightened the demand for real-time analytics and adaptive business strategies. Selecting an optimal stream processing tool is critical, yet challenging, due to the wide array of available platforms and the complexity of requirements in modern e-commerce environments. This thesis addresses the gap by applying a structured decision-making framework, based on the Analytic Hierarchy Process (AHP), to guide e-commerce organizations in evaluating and selecting stream processing tools aligned with their operational and strategic needs.

The research employs a multi-method case study within an European e-commerce company, combining qualitative data from stakeholder interviews, documentation analysis, and observations, with quantitative pairwise comparisons to establish and weight key selection criteria. Six stream processing platforms: Apache Flink, Apache Spark Structured Streaming, Apache Kafka Streams, Apache Storm, Apache Samza, and Google Cloud Dataflow are systematically evaluated against criteria such as fault tolerance, performance, state and event handling, integration, operability, and cost within a dynamic pricing case study. The findings demonstrate how a criteria-driven methodology can support organizations in making informed and context-aware technology choices.