Single cell transcriptomics has emerged as the preferred tool to define cell identity through the analysis of gene expression signatures and has been pivotal to many recent scientific breakthroughs. The development of microfluidic-based systems coupled with cell and transcript barcoding has allowed the broad scientific community access to high throughput single cell RNA-seq. However, there is limited information comparing the performance of different scRNAseq systems in biologically relevant complex tissues. Here, we present a systematic comparison on the capacity of three well established and widely use microfluidic-based 3’-scRNAseq platforms (Drop-seq, inDrop and 10X) to analyse tumours that present high cell diversity. Our experimental design includes samples from the same tumours across all platforms, which provides a comparable dataset to examine clustering performance, cell type assignment, and their potential to interrogate biologically relevant molecular pathways. Additionally, we experimentally model technical noise and the decay on the quality of scRNAseq data in a longitudinal temporal analysis of the same sample. We determined that all platforms were able to cluster the data identifying the major distinct cell types, however, we found cell-type-associated biases driven by a differential gene identification sensitivity and technical noise. In this study, we describe how platform-specific biases can impact on the biological insights driven from the data. In conclusion, besides the economical value and accessibility of each platform, potential biological biases and technical noise should be considered for future scRNAseq applications.