Expression data from RT-qPCR (reverse transcription quantitative PCR) needs to be normalized to account for experimental variability among samples caused by differential yields of the transcripts in RNA extraction or in the reverse transcription. The most common method is to normalize against one or more reference genes (RG). We have selected RGs suitable for normalization of RT-qPCR raw data in Saccharomyces cerevisiae during alcoholic fermentation. The RGs were evaluated by three different statistical methods. The suitability of the selected RG sets was compared with ACT1, a commonly used non-validated single RG, by normalizing the expression of two target genes. Expression profiles of the target genes revealed the risk of misleading interpretation of expression data due to an unreliable RG.