Serial electron crystallography has been developed as a fully automated method to collect diffraction data on polycrystalline materials using a transmission electron microscope. This enables useful data to be collected on materials that are sensitive to the electron beam and thus difficult to measure using the conventional methods that require long exposure of the same crystal. The data collection strategy combines goniometer translation with electron beam shift, which allows the entire sample stage to be probed. At each position of the goniometer, the locations of the crystals are identified using image recognition techniques. Diffraction data are then collected on each crystal using a quasi-parallel focused beam with a predefined size (usually 300-500nm). It is shown that with a fast and sensitive Timepix hybrid pixel area detector it is possible to collect diffraction data of up to 3500 crystals per hour. These data can be indexed using a brute-force forward-projection algorithm. Results from several test samples show that 100-200 frames are enough for structure determination using direct methods or dual-space methods. The large number of crystals examined enables quantitative phase analysis and automatic screening of materials for known and unknown phases.