To characterize how technical variables affect the reliability of high-throughput RNA sequencing (RNA-seq), the SEQC/MAQC-III Consortium carried out RNA-seq using different sequencing platforms and protocols across different geographical laboratory sites. From the >100 billion sequence reads generated, they compared the ability of RNA-seq procedures to identify known features of the RNA samples relative to each other and relative to gene expression microarrays and quantitative PCR. Overall, good accuracy and reproducibility were seen across approaches, platforms and locations when identifying differentially expressed genes (if appropriate bioinformatic filters were used). However, RNA-seq and microarrays showed inaccuracies for the quantification of absolute RNA levels, and all approaches had gene-specific biases. In parallel, Li et al. show that RNA-seq platforms differ in their ability to detect RNA variants, such as splice site isoforms.