A common theory around here is 10%, I would buffer that to 20 - 25%.
Diversity in online markets doesn't necessarily mean product/service diversity, and this can lead to duplicate content.
On pages that are (IMO less than 20% similar) you do have 2 courses of action to avoid the Google wraith.
Crosslinking - can to bad again if you over do it, but their are significant benefits, particularly when used to avoid duplication.
An example: news releases. As all sites are owned by the same client the corporate news is likely to be the same in all sites.
Only one site should retain the news pages and all other sites should link to those pages. Should the design layout of each individual site be the same the graphical imagery of the interface (shell) can also link to the appropriate design by using multiple cascading styles sheets on the primary web pages, serving up the appropriate style on where the request is coming from (one domain or another). If design integrity (look and feel) is not that important this makes your job easier.
Unique elements, tags, and attributes in each domain also varies content and this can actually be close to 10% depending on how innovative you are.
One example used for a client (earth science) is Plate Tectonics and Tectonic Plates.
The scientific community only uses Plate Tectonics as this define the actual science.
The general population not being as scientific savvy as scientists commonly refer to Tectonic Plates.
Quite a nice market diversity, and a beautiful way of targeting normally duplicate pages on different sites in the text content, alt tags, titles (elements and attributes) as well as file names (images, pages, and objects) and directory names.
In this case it is good to remember that "content" is not reserved for just "text".