Person re-identification (Re-ID) aims to match pedes- trians under dis-joint cameras. Most Re-ID methods for- mulate it as visual representation learning and image search, and its accuracy is consequently affected greatly by the search space. Spatial-temporal information has been proven to be efficient to filter irrelevant negative sam- ples and significantly improve Re-ID accuracy. However, existing spatial-temporal person Re-ID methods are still rough and do not exploit spatial-temporal information suffi- ciently. In this paper, we propose a novel Instance-level and Spatial-Temporal Disentangled Re-ID method (InSTD), to improve Re-ID accuracy. In our proposed framework, per- sonalized information such as moving direction is explic- itly considered to further narrow down the search space. Besides, the spatial-temporal transferring probability is disentangled from joint distribution to marginal distribu- tion, so that outliers can also be well modeled. Abun- dant experimental analyses are presented, which demon- strates the superiority and provides more insights into our method. The proposed method achieves mAP of 90.8% on Market-1501 and 89.1% on DukeMTMC-reID, improv- ing from the baseline 82.2% and 72.7%, respectively. Be- sides, in order to provide a better benchmark for per- son re-identification, we release a cleaned data list of DukeMTMC-reID with this paper: https://github. com/RenMin1991/cleaned-DukeMTMC-reID/