Diversifying Science - Is it as simple as replicating 'programs that work'?

What's so hard about designing and running an effective STEM diversity program? Ask any director about his/her program or read any program proposal outlining what the PI plans to do, and you'll find their 'lists' to be remarkably similar. Comprehensive support, early exposure to research, selecting the best students, developing community, meeting financial need, faculty mentoring, and an emphasis on excellence make up the typical short list of requisites to build an effective program.

However, in spite of this consensus about what to do and after 40 years of diversity programs supported by billions of public and private dollars, under-representation persists. How can we more effectively help our under-represented (and all) students succeed in STEM majors and related careers?