The identification of single microorganism in food samples by conventional plating techniques or molecular genetic methods requires a time consuming enrichment step. Raman spectroscopy in combination with a suitable extraction method however offers the possibility to rapidly identify bacteria on a single cell level. Here we evaluate the two well-known bacteria extraction methods from milk: "buoyant density centrifugation" and "enzymatic milk clearing" towards their recovery efficiency and their compatibility with Raman spectroscopy for a rapid identification of microorganisms in milk. The achieved recovery yields are slightly better compared to those which are already applied for food investigations, where a loss of one order of magnitude is usually reached. For example, buoyant density centrifugation allows collecting up to 35% of the milk-spiked microorganisms. To prove the suitability of the isolation techniques for use in combination with the spectroscopic approach, a small Raman database has been created by recording Raman spectra of well-known contaminants in dairy products. Two subspecies of Escherichia coli and three different Pseudomonas species, which were inoculated to UHT (ultra-high-temperature processed) milk and afterwards extracted by the two techniques mentioned above, were analysed. At a first glance, grave spectral artefacts caused by the matrix itself or especially by the extraction techniques were not obvious. But via chemometric analysis, it could be shown that these factors noticeably influence the identification rates: while the samples prepared via milk clearing did not provide sufficient identification results, buoyant density centrifugation allows an identification of the investigated species with an overall accuracy of 91% in combination with linear discriminant analysis.