This chapter explores recent methodologies leveraging small datasets for the optimisation of material properties in chemical processes. By focusing on vanadium catalysts, we demonstrate how data-driven approaches can effectively help catalyst design and reaction optimisation, even with limited experimental data. The aim is to illustrate the potential of small data methodologies in addressing the complexities and variabilities inherent in traditional large dataset approaches. The advantages of using small datasets are disclosed, but we also address the associated disadvantages, providing a comprehensive analysis of how limited datasets can drive significant advancements in catalyst performance. Using the case study of vanadium-catalysed epoxidation reactionsEpoxidation reactions, we emphasise the strategic use of small datasets in chemical research, underscoring the importance of both quantitative and qualitative dataQualitative data. The demonstrated methodology involves the meticulous collection, digitisation, and standardisation of reaction data, transforming it into a machine-readable format using standardised chemical descriptorsChemical descriptors and SMILESSMILES strings. By deploying machine learningMachine learning in chemistry techniques, the analysis uncovers critical patterns and relationships within the dataset, identifying key chemical features that significantly influence epoxidation reaction yieldsReaction yields. Despite the current challenges of working with small datasets, the insights gained from this approach have profound implications for interpreting chemical properties, driving more efficient and sustainable processes.