Lesotho has customised and deployed OpenLMIS as a national electronic logistic management information system (eLMIS) to improve visibility and availability of health commodities across all levels of the supply chain. However, during rollout, key requisitions and stock management workflows faced timeouts and system crashes that resulted in user frustration and therefore negatively affecting the envisaged benefits. This paper describes a multi-layer performance optimisation of the Lesotho eLMIS focusing on application logic, cross-microservice communication, caching, database access and deployment topology. Profiling was used to locate performance bottlenecks such as fragmented GET-based requests, repeated retrieval of static reference data and N+1 stock on hand database queries. To improve the eLMIS performance, POST-based endpoints, Redis-based caching, batched database queries, as well as tuning of the database engine and deployment environment were implemented. As a result, requisition and stock management workflows saw a performance gain of at least 900% and 600% respectively. The approach provides a practical performance engineering model for similar OpenLMIS-based deployments.