Developing a Data-Driven FinOps Framework for Automated Cloud Cost Allocation and Variance Analysis in Hybrid-IT Environments using Machine Learning

The financial governance of contemporary enterprise infrastructure increasingly takes place in Hybrid-IT environments, where on-premises systems coexist with workloads distributed across multiple public-cloud providers, producing cost data under fundamentally different logics that resist unified analysis. The FinOps discipline has emerged as the institutional response, and the FinOps Open Cost and Usage Specification (FOCUS) has been proposed as a data standard to bridge the comparability gap; however, empirical evidence on how Machine Learning techniques, FOCUS-aligned data, and practitioner adoption interact in Hybrid-IT settings remains limited. This dissertation addresses that gap through a mixed-method sequential explanatory study comprising three empirical strands: a controlled Machine Learning experiment applying Random Forest and Bidirectional LSTM models to a FOCUS-aligned Hybrid-IT dataset combining the public FOCUS Sample Data with a synthetic on-premises dataset of 510 servers over 48 months; an online survey of 28 FinOps practitioners (89 per cent in Hybrid-IT environments); and five semi-structured interviews with senior practitioners and standard-setters including the FOCUS architect at the FinOps Foundation. Random Forest achieved R² = 0.98 for cost allocation and BiLSTM achieved MAPE = 4.00 per cent on the 48-month forecasting task, demonstrating that FOCUS-aligned data can support high-accuracy automated cost analysis. The survey produced nine substantive findings including a 27 per cent FOCUS awareness gap, a statistically significant awareness–belief correlation (Spearman rho = 0.535, p = 0.005), and a trust-led cluster of adoption barriers. The interviews surfaced six themes, most notably the cross-cultural convergence that cost allocation is fundamentally a financial rather than a technical decision. The combined evidence supports a defensible claim: data standardisation through FOCUS is a necessary but not sufficient condition for the responsible adoption of Machine Learning in Hybrid-IT cost management.