Abstract
Process-control systems in industrial and infrastructure sectors are critical for operational efficiency, yet robust methodologies for evaluating their long-term performance gains in developing economies are lacking. This gap hinders evidence-based investment and optimisation. This case study aims to develop and apply a novel time-series forecasting model to quantify efficiency gains from process-control system implementations. The objective is to provide a replicable methodological framework for performance evaluation. A comparative case-study analysis was conducted using longitudinal operational data from multiple sites. The core methodological innovation is a hybrid forecasting model integrating an ARIMA component with an intervention analysis term, formalised as $Yt = \mu + \phi Y{t-1} + \theta \epsilon{t-1} + \omega It + \epsilont$, where $It$ is a step function for system implementation. Model parameters were estimated using maximum likelihood, and forecast uncertainty was quantified with 95% prediction intervals. The model forecasts a sustained 18.5% aggregate improvement in system throughput efficiency over the forecast horizon post-intervention. Statistical inference indicates this gain is significant (p < 0.01), with model diagnostics confirming stationarity in the forecast residuals. The proposed time-series model provides a statistically rigorous framework for attributing efficiency improvements to process-control interventions, moving beyond descriptive assessment. Adopt the hybrid forecasting model for baseline efficiency measurement and post-implementation audits. Engineers and planners should integrate such models into the project lifecycle to validate control-system ROI. process control, time-series analysis, forecasting, efficiency measurement, intervention analysis, infrastructure systems This paper introduces a novel hybrid time-series model for quantitatively isolating and forecasting the efficiency gains attributable to process-control system upgrades, demonstrated with longitudinal data.