Increasing demand of sustainable energy utilization boosts the development of high performance energy storage technologies in particular lithium-ion batteries [1]. Transition metal oxides have attracted tremendous attention as promising electrode materials for lithium-ion batteries due to their abilities to deliver much higher specific capacities at a relatively inexpensive cost [2]. For structural considerations, it is widely acknowledged that the ideal layered or spinel structures are crucial for reversible lithium insertion and extraction as they can provide space for efficient ionic transportation while maintain good stability of oxygen-metal frameworks [3]. It is also well known for many metal oxides that phase conversion reactions can occur upon a deep lithium insertion, which will increase the overall battery capacity along with severe structure modifications such as collapsing single-crystalline particles into small nanoparticle composites [1, 2]. Specifically for spinel iron oxide (Fe3O4 or magnetite), multiple phase transformation along with structure changes have been previously found by static characterizations after complete relaxation from electrochemical reactions [4, 5]. However, real batteries do not operate under true equilibrium conditions, which may make actual phase transitions carry out in different and complicated ways. Thus, it is of great importance to understand the structural evolution throughout a full lithiation process so as to precisely regulate the lithiated phases with desired structures for future design and applications of these electrode materials. For such purposes, we need to utilize in situ transmission electron microscopy (TEM) and spectroscopy [6], which have sufficient spatial resolution and chemical detectability to obtain direct correlations between microstructure and electrochemistry. In this work, we implement the in situ dry-cell setup for real-time observation of electrochemical lithiation inside a TEM (JEOL 2100F operated at 200 kV), where the composite electrode containing Fe3O4 nanocrystals and the amorphous carbon support will be lithiated, while extra side reactions are eliminated with no participation of liquid electrolyte [6]. Using a strain-sensitive scanning transmission electron microscopy (STEM) approach, we were able to observe the phase changes within a single Fe3O4 nanocrystal, and clearly differentiate two reaction steps: 1) lithium intercalation with structural change from spinel to rocksalt by displacing the Fe cations from tetragonal 8a sites to the nearby octahedral 16c sites; and 2) the subsequent conversion to form composites of ultrafine Fe nanoparticles along with amorphous Li2O. With real-time observation, we found that these two steps are largely overlapped due to the kinetic effects under nonequilibrium electrochemical conditions. The results are consistent with synchrotron x-ray diffraction and well explained by calculations using both first-principle DFT and phase-field theory. Our findings provide insights into understanding lithiation mechanisms in spinel structures, and also show implications for improving performance in future design of battery electrodes.