Plant phenomics is becoming a common tool employed to characterize the mode of action of biostimulants. A combination of this technique with other omics such as metabolomics can offer a deeper understanding of a biostimulant effect in planta. However, the most challenging part then is the data analysis and the interpretation of the omics datasets. In this work, we present an example of how different tools, based on multivariate statistical analysis, can help to simplify the omics data and extract the relevant information. We demonstrate this by studying the effect of protein hydrolysate (PH)-based biostimulants derived from different natural sources in lettuce and tomato plants grown in controlled conditions and under salinity. The biostimulants induced different phenotypic and metabolomic responses in both crops. In general, they improved growth and photosynthesis performance under control and salt stress conditions, with better performance in lettuce. To identify the most significant traits for each treatment, a random forest classifier was used. Using this approach, we found out that, in lettuce, biomass-related parameters were the most relevant traits to evaluate the biostimulant mode of action, with a better response mainly connected to plant hormone regulation. However, in tomatoes, the relevant traits were related to chlorophyll fluorescence parameters in combination with certain antistress metabolites that benefit the electron transport chain, such as 4-hydroxycoumarin and vitamin K1 (phylloquinone). Altogether, we show that to go further in the understanding of the use of biostimulants as plant growth promotors and/or stress alleviators, it is highly beneficial to integrate more advanced statistical tools to deal with the huge datasets obtained from the -omics to extract the relevant information.