Various things:

• uncompressed audio (perhaps needlessly, for lower-end CPUs)
• pre-rendered cutscenes, multiple resolutions
• large texture sizes (also think normal maps etc), multiple resolutions
• redundant or unused data
• large model sizes, or more assets in general
• previous issues may be multiplied by multi-language support

I would also say it could be a problem of art direction, having no constraint and ignoring smart techniques that use less data.

Also, a higher-budget --> more preorders?
More data --> more game?