Customer Engagement Scoring with Machine Learning: Integrating Clickstream, Social, and CRM Data

Abstract

This study investigates how machine learning is used to construct customer engagement scores from clickstream, social media, and CRM data in recent marketing research. It asks how engagement is conceptualized and operationalized, which data sources and feature representations are employed, and to what extent multi source integration is implemented in existing models. Using a systematic literature review of peer reviewed studies from 2017 to 2022, the article maps contexts, data characteristics, modelling approaches, and evaluation practices. The synthesis shows that most studies model engagement as a single channel behavioral outcome and rely on task specific indicators, with deep learning and tree-based methods frequently outperforming traditional models. Empirical work that jointly exploits clickstream, social, and CRM data in unified engagement scoring architectures remains limited. The review develops an integrative view of current design patterns, identifies conceptual and methodological gaps, and outlines priorities for future research on cross channel, value-oriented engagement scoring with machine learning.