The process of reverse engineering using 3D scanning can yield many outputs and there is certainly some confusion between them. I hope this brief explanation of options can help set you down the correct path for your needs.
The term raw is certainly a defining feature. Nearly all 3D acquisition devices capture surface geometry in its purest form, a series of points defined by specific 3D coordinate. The position of these points relative to each other represent the surface of the geometry being acquired. Their positional accuracy is determined by the acquisition hardware being used. It is the most accurate representation of the surface, with a caveat. Raw data is as it sounds, the good and the bad. Be it structured light or laser based scanning hardware, reflections may occur resulting in false feedback to the hardware sensors. These result in residual noise scattered along with the remaining good points that were captured. With raw data also comes larger file sizes, containing all the mathematical positions of millions of points. Long range applications, including Lidar, time of flight and phase shift systems can capture millions of raw points in a matter of seconds. Also, anyone who has opened a raw point cloud knows that they are harder to visualize compared to other data sets that will soon be mentioned. Common formats are .XYZ, .ASC & .TXT.