The video demonstrates adding footage to camera tracker, then auto-tracking, then solving for 3D space. The 3D solver basically tries to understand the angle of the camera. They draw a triangle in relation to the auto-tracked points, which is to help solve so it knows what orientation is UP (Y). This reorients the tracking points and camera so the plane will more closely match the 3D space. They only do Y because that is the only axis they are concerned with here. They are not being super careful to make the orientation perfect.
They add a position constraint to the motion tracker, which indicates where the 0,0,0 coordinates are, so when they add the plane it is very close to where it should be.
They go on to make a plane for the statue to sit on. This is only to catch the shadow of the statue. They add the footage to it, and mark it as compositing background with no self-shadowing. This makes it blend into the tracked footage, while also being able to render the shadow.
They add a background object and copy the footage material tag from the plane. This is so the footage is rendered behind the plane and statue.
They delete the motion tracker because they already solved for tracking and 3D space, which created the solved camera automatically. You don't need to delete the tracker if you don't want to, and I recommend that you don't in case you want to re-do tracking / solving.
They add a sky object and add a copy of the footage material with luminance turned on. This utilizes the footage as the light source for the render. They tweak it by also setting exposure to 1. They add compositing tag to sky and unmark seen by camera, otherwise it would override the background at render.
They then add a few rendering options for GI, color correction, and denoiser.
Camera is adjusted for exposure settings. You will have to play to find best setting. They already know what will work best for the example.
Then they finally place the statue and match the orientation of the shadow catcher plane, then render.
Vizn's post in Rendering a voronoi fracture object in the camera tracker was marked as the answer