Journal Design Engineering Masthead
African Structural Engineering | 02 October 2003

A Quasi-Experimental Design for Cost-Effectiveness Analysis of Industrial Machinery Fleet Diagnostics in South Africa.

P, i, e, t, e, r, v, a, n, d, e, r, M, e, r, w, e, ,, T, h, a, n, d, i, w, e, N, k, o, s, i
Quasi-experimental designCost-effectiveness analysisIndustrial machineryPredictive maintenance
Difference-in-differences framework isolates causal effect of predictive diagnostics
14.2% cost reduction achieved in treatment fleet versus matched control
Method provides rigorous counterfactual analysis for maintenance investments
Evidence supports cost-effectiveness in resource-constrained industrial settings

Abstract

{ "background": "The management of industrial machinery fleets in resource-constrained environments requires robust, evidence-based strategies to optimise maintenance expenditure and operational availability. Current evaluations often lack rigorous counterfactual analysis, making true cost-effectiveness difficult to ascertain.", "purpose and objectives": "This case study aimed to develop and apply a quasi-experimental design to measure the cost-effectiveness of a novel predictive diagnostics system implemented for a large haul truck fleet in the mining sector. The primary objective was to quantify the net financial impact against a comparable control group.", "methodology": "A difference-in-differences (DiD) framework was employed, comparing maintenance cost trajectories between a treatment fleet (\(n=42\) vehicles) fitted with advanced telematics and a matched control fleet (\(n=38)\) using legacy scheduled maintenance. The core statistical model was $Y{it} = \\beta0 + \\beta1 \\text{Treat}i + \\beta2 \\text{Post}t + \\delta (\\text{Treat}i \\cdot \\text{Post}t) + \\epsilon_{it}$, where $\\delta$ captures the causal effect. Inference was based on cluster-robust standard errors at the fleet level.", "findings": "The diagnostic intervention yielded a statistically significant reduction in mean monthly maintenance costs per vehicle. The DiD estimator, $\\delta$, was -ZAR 18,750 (95% CI: -ZAR 22,100, -ZAR 15,400), representing a 14.2% cost reduction relative to the control group's post-intervention mean. The analysis showed a high likelihood (p < 0.01) that the observed savings are attributable to the new system.", "conclusion": "The quasi-experimental design provided a rigorous, defensible method for isolating the causal effect of the diagnostics system on maintenance costs. The results demonstrate that targeted predictive technologies can be cost-effective in heavy industrial applications within the regional context.", "recommendations": "Fleet