Where and When: How Continuous Lift Monitoring Answers the Two Questions That Drive Service Efficiency

Abstract

Effective lift service has always depended on two questions: where is the problem, and when did it start? In the absence of continuous monitoring, answers to both questions arrive late, filtered through tenant callbacks, manual inspection logs, and technician recall. This paper examines how a controller-agnostic IoT platform, deployed across a heterogeneous fleet of commercial, transit, and healthcare lift installations, changes the answer to both questions in practice. Central to the approach is the absence of modelled baselines. Instead, two algorithms based on isolation forests and adaptive windowing are presented which detect anomalies by different methodologies to cover different kinds of anomalies. This approach is contrasted with fixed global threshold alerting, which the authors implemented and subsequently replaced following significant false positive rates. Operational case studies illustrate the practical outcomes. The paper discusses the known limitations of adaptive baseline methods, including the initialisation period required before reliable detection is possible, and the latency in alert resolution once a fault has been cleared.