Interpretable Models over Customer 360 Data for Stakeholder Trust and Governance

Abstract

Organisations in many sectors increasingly rely on integrated customer data platforms that consolidate operational, transactional, behavioural, and third-party information into so-called Customer 360 representations. These holistic profiles enable predictive and prescriptive analytics for marketing, service, risk, and compliance use cases. At the same time, regulatory expectations, internal model risk standards, and public scrutiny place growing emphasis on interpretability, governance, and demonstrable fairness of data-driven decisions derived from such systems. This paper examines interpretable modelling approaches over Customer 360 data with a focus on their suitability for supporting stakeholder trust and governance practices across business, technical, and oversight functions. The discussion characterises Customer 360 data along dimensions of heterogeneity, temporal structure, sparsity, and data quality, and analyses how these characteristics interact with transparency requirements and explanation methods. The paper surveys model classes that are inherently interpretable, as well as post-hoc explanation techniques, and assesses their strengths and limitations under typical Customer 360 workloads, including propensity scoring, retention prediction, next-best-action ranking, and early-warning signals. Particular attention is given to the alignment between model explanations, organisational decision processes, and the information needs of different stakeholders such as product owners, legal and compliance teams, auditors, and affected customers. The paper further sketches an architectural perspective on integrating interpretable models into Customer 360 platforms, touching on data lineage, policy enforcement, monitoring, and documentation artefacts. An evaluation framework based on quantitative and qualitative criteria is proposed to reason about model performance, stability, comprehensibility, and governance readiness in a coherent manner. The paper concludes with observations on practical trade-offs and open questions in operationalising interpretable Customer 360 modelling at scale.