A Comparative Literature Review of Data-Driven Decision-Making Phases in Information Systems: Toward an Integrated Six-Phase Framework

Abstract

This study presents a comparative literature review on data-driven decision-making (DDDM) within Information Systems (IS) and synthesized an integrated framework including six phases to reduce conceptual fragmentation in existing research. Studies published between 2016 and 2025 were systematically reviewed, mapped, and compared, with the evidence consistently presented across six phases: data collection, data preprocessing, feature selection, model development, model evaluation, and decision support. The review is guided by three research questions. RQ1 examines how data quality, governance, privacy and security, and organizational readiness are addressed in the early stages. RQ2 analyses dominant approaches to complexity reduction and model development, highlighting trade-offs among performance, interpretability, and organizational applicability. RQ3 evaluates how reliability is reported and how analytical outputs are translated into actionable decisions, with particular attention to Explainable AI (XAI), Human-in-the-Loop (HITL), and ethical considerations. The findings indicate that key challenges include limited organizational capacity, concerns related to privacy and security, and inconsistencies in data quality. Critical enablers include robust governance structures, analytics literacy, leadership support, and a strong data-driven culture. The proposed framework provides a unified reference model for comparing methods, clarifying inter-phase relationships, and improving transparency and reproducibility in DDDM processes within IS. To support validation against state-of-the-art approaches, the framework is compared with existing pipelines, highlighting differences in governance integration, XAI/HITL incorporation, and controllability mechanisms. Finally, the framework is demonstrated using a case study based on the BTS flight performance dataset, illustrating phase-by-phase implementation and the translation of model outputs into explainable decision support.