Bikesharing has gained popularity over the years and now faces the challenge to be responsive and meet the growing demand. Thus, understanding the temporal pattern of bikesharing trips is paramount. This study examined data from the Bikesampa bikesharing system (a fixed station system operating in the Brazilian city of São Paulo).First, a k-means clustering was applied to group bikesharing stations according to hourly demand of bike pick-ups and returns. The results revealed three clusters of stations with well-defined and distinct temporal patterns: (i) balanced, (ii) unbalanced, with higher rates of bike pick-up in the morning and (iii) unbalanced, with higher rates of bike return in the morning. A spatial autocorrelation analysis showed that stations belonging to each cluster are not randomly distributed over space, indicating that the system may require different rebalancing strategies depending on the area where stations are located and suggesting the association with built environment characteristics around the stations. This hypothesis was confirmed through a probit model, which shows the association between the temporal demand patterns and operational, spatial, and socioeconomic attributes. Such understanding can help guiding the development of operational strategies and user incentive policies to improve the efficiency of bikesharing systems. Also, it allows for the prediction of temporal patterns when analyzing new strategies regarding bicycle repositioning or implementation of stations in the short and medium term, or after socio-spatial changes in the long-term.