We examine statistically the duration of substorm growth phases which occurred during the year 2010. We find that there is no strong link between the duration of a growth phase and the average dayside reconnection rate during that growth phase. Rather, there is a distribution of growth phase durations between 10 mins and 3 hours, with a mean of 80 mins. We infer that substorm onset occurs randomly within this distribution, irrespective of the level of solar wind driving. Such a “typical’ growth phase duration explains why the open magnetic flux content of the magnetosphere at substorm onset depends on average on the dayside reconnection rate. We also infer that the rate of nightside reconnection during the expansion phase closely matches the dayside rate at the time of substorm onset. This leads to a substorm intensity that is related to the open flux at onset, as has been observed in previous studies. We suggest that pseudo-breakups are failed substorm onsets and that the growth phase continues until one of these attempts to initiate nightside reconnection succeeds.