In “relative scale” scenes, the scale of the scene is determined by the initial height (Y-value) of the camera in the scene. Lower camera heights result in larger content. Higher camera heights scale the content smaller. The initial height of the camera translates to the distance to the surface. As an example, if you place your camera at an initial height of 1 in your scene, then 1 unit represents the distance between your phone and the surface in front of you. If you 3D model with a height of 1, it will appear to be as tall as the distance from your camera to the surface in front of you. If then change the initial camera height to 2, and reload the page, the object would be half the size.
You can reference more in the tutorial video here: https://www.youtube.com/watch?v=bfQakA5-TQI&t=43s