Resistive Random Access Memory (RRAM) has seen significant developments in the past years. An important improvement is to reduce the device-to-device variability that limits RRAM implementation. A proposed approach is to measure a high number of devices to statistically identify trends and evidences behind the cause. Here, we present a method to characterise the RRAM switching behaviour and analyse a large dataset of SnOx RRAM, in a statistical fashion for further optimisation and modelling.