In recent times we have seen the development of many “-omics” technologies. One of the youngest is undoubtedly metabolomics, which aims to define the whole chemical fingerprint unique to each specific organism. The development and optimisation of an untargeted high-throughput method capable of investigating the volatile fraction of a biological system represents a crucial step for the success of such holistic approaches, and specific optimisation criteria must be developed in connection with suitable experimental designs. In this paper experimental designs (D-optimal) were applied for the first time as an automatic optimisation tool to an untargeted HS-SPME-GC-TOF method. In this case, optimal conditions correspond to a maximal number of detected features, in order to provide a fingerprint that is as complete as possible. The system under study is the grape berry. Four variables were considered: the type of fibre, extraction time, equilibration time and temperature. The results show that the D-optimal design methodology provides an easily interpretable assessment of experimental settings. This and other specific properties of the D-optimal design, such as the possibility to explicitly exclude certain experimental conditions, make it an extremely suitable strategy for method optimisation in untargeted metabolomics.