The variables affecting the direct matrix assisted laser desorption ionization mass spectrometry-based analysis of wine for classification purposes have been studied. The type of matrix, the number of bottles of wine, the number of technical replicates and the number of spots used for the sample analysis have been carefully assessed to obtain the best classification possible. Ten different algorithms have been assessed as classification tools using the experimental data collected after the analysis of fourteen types of wine. The best matrix was found to be α-Cyano with a sample to matrix ratio of 1:0.75. To correctly classify the wines, profiling a minimum of five bottles per type of wine is suggested, with a minimum of three MALDI spot replicates for each bottle. The best algorithm to classify the wines was found to be Bayes Net.