There is usually a driving reason for the production of visualizations, if for no other than to 'see' the design, or show it to other planners. The reason may be to explain an alignment issue, to make a particular design decision, or explain the proposed functionality of the design. The particular message or issue for which the visualization is being produced should drive decisions like level of detail in the model, level of 'realism' required, what type of views will be produced, and camera location. As an example, an overhead aerial view of a model would be more useful for explaining layout or planning issues, where an eye-level view would give a more experiential view of the contextual influence.
To explain function and use, a model view may need to be linked to other analytical data, and include moving cars or annotation. A view intended to show 'what it will be like' will need more project details and contextual information like trees, people or realistic cars. An important decision to be made is just how much of the project needs to be fully modeled in 3D. Modeling only those elements needed to 'tell the story' can save time and effort. If the aesthetic treatment for a bridge is the issue, for instance, one need only model and detail the bridge elements, and leave out or simplify the detail of the adjacent roadway elements.