A scene graph represents objects and their spatial relationships as a hierarchy rather than as one flat list of world-space coordinates.

Each node stores a transformation relative to its parent. The world transform of a child is obtained by composing transforms along the path from the root:

$$M_{world}=M_0M_1\cdots M_k.$$

This makes articulated or nested structures natural. A wheel can be positioned relative to a car, and the car relative to the world. Moving the car then moves every descendant without rewriting their local geometry.

Hierarchies also separate reusable geometry from instances. Several nodes can refer to the same mesh while carrying different transforms or material state.

The tree structure encodes ownership of transformations, not necessarily physical attachment. A poor hierarchy can therefore make independent motion difficult even when the rendered image initially looks correct.

Scene graphs are a reusable way to organize transform composition, animation and grouped objects.