Updated project metadata. Computing absolute protein abundances using mass spectrometry (MS) is a widely used technique in quantitative biology. An important and often overlooked aspect in this methodology is to assess technical reproducibility, i.e. how precise are predictions of abundance when we use the instrument on repeated occasions to measure the same sample. Here, we present a proteomics dataset of Saccharomyces cerevisiae with both biological and inter-run technical triplicates, which we use to analyze both accuracy and precision of the MS instrument. We also investigate how we can improve the quality of predictions by using 4 alternative methods for estimating absolute protein abundance starting from MS intensities. We show that using a simple normalization and rescaling approach performs equally accurate, but much more precise, than methods that rely on external standards. Furthermore, we show that technical reproducibility is significantly lower than biological reproducibility for all the evaluated methods. The results presented here serve as a benchmark for assessing the best way of interpreting MS results to compute protein abundances, and as a consideration of the limitations of the technique when interpreting results.