Oral cancer has a poor prognosis of only 50% even in the light of current technological advances. This poor prognosis may be attributed to the still unmet clinical need to diagnose oral pre-cancer and dysplasia. Raman spectroscopy, which can detect subtle biochemical changes, has been explored for the diagnosis of cancer. This study aims to address the clinical need by exploiting the high amplification factor of Surface Enhanced Raman Spectroscopy (SERS) to analyse the saliva samples of 10 healthy controls and 10 patients with oral dysplasia. Furthermore, this technique was compared to conventional Raman spectroscopy. The saliva samples were centrifuged at 14000g for 15 minutes and the supernatant was applied directly on the SERS substrate and dried. Simultaneously, the saliva samples were prepared in the same way on slides for conventional Raman analysis. A peak at 2108 cm-1, attributed to salivary thiocyanate was present in all samples from dysplasia subjects but absent in samples from healthy non-smoking subjects. Partial least squares – discriminant analysis models for classification of oral pre-cancer were developed for both Raman spectroscopy and SERS to discriminate between healthy, mild and moderate dysplasia cohorts.