With most lidar systems, an initial yaw (or heading) is determined at the beginning of the data set, and from there on out the IMU maintains this initial yaw using it's gyroscope data. This initial yaw determination is referred to as the Kinematic Alignment:

During a kinematic alignment, the vehicle moves in a straight line. GNSS positions recorded on this straight line form a course over ground, which itself is a direction of travel in world space (south, southeast, etc.). Thus, during the kinematic alignment the IMU's initial heading is initialized as this course over ground. 

Typically kinematic alignments are performed at a fast speed, faster than the vehicle speed during the mapping portion of the mission. This is because the INS, and trajectory processing softwares like InertialExplorer and NavLab, have minimum speed parameters. The rationale behind these minimum speed thresholds is that you want to align at a good time, i.e. with a clear view of the sky and right before beginning your mapping mission, thus a minimum speed requirement allows the system to wait, unaligned, until the operator is ready to begin the mission. 

Without a proper kinematic alignment maneuver, the trajectory solution will likely exhibit heading or yaw errors, which will in turn result in relative accuracy issues in pointcloud, particularly with vertical features.