Abstract
Evaluating the operational efficiency of water treatment infrastructure in developing nations is critical for public health and resource management. However, rigorous, field-based comparative analyses of system performance are scarce, particularly those employing robust experimental designs to isolate causal factors. This study aims to methodologically evaluate and compare the efficiency gains of different water treatment facility systems. Its primary objective is to quantify the causal impact of technological and managerial interventions on system performance metrics. A quasi-experimental, difference-in-differences design was employed, analysing longitudinal performance data from multiple treatment facilities. The core statistical model is $Y{it} = \beta0 + \beta1 \text{Treat}i + \beta2 \text{Post}t + \delta (\text{Treat}i \times \text{Post}t) + \epsilon{it}$, where $Y{it}$ is the efficiency outcome. Inference is based on cluster-robust standard errors to account for facility-level heterogeneity. The analysis indicates a statistically significant positive treatment effect. Facilities implementing integrated membrane filtration with enhanced coagulation protocols demonstrated a mean efficiency increase of 22.4% (95% CI: 18.1, 26.7) in contaminant removal rates compared to control groups using conventional sedimentation. The quasi-experimental framework provides robust evidence that specific technological upgrades, when coupled with structured operational protocols, yield substantial and measurable improvements in treatment efficiency. Infrastructure investment should prioritise the phased adoption of integrated membrane systems. Policy must concurrently support targeted operator training programmes to ensure sustained operational gains. water treatment efficiency, quasi-experimental design, difference-in-differences, infrastructure diagnostics, process engineering This paper provides a novel application of a causal inference framework for comparative infrastructure diagnostics in a resource-constrained context, generating a validated longitudinal dataset for future modelling.