Can you really control diversity outcomes?
You have NO CONTROL!!! I have bad news. Now that you have one detailed problem statement and you understand your retention data, you don't control retention. This is a common pitfallll - believing that you can control what you're measuring. Even though you don't control the retention, you can control the inputs to retention. But, this requires that you know what influences retention.
I call these influencing factors "inputs". Here are some examples of controllable inputs that can be measured: # of mentoring relationships with Black engineering students, # of black engineering students recruited, # of Black engineering graduates hired, # of Black engineering students sponsored, etc. There will be several inputs to Black retention, but you are looking for those inputs that influence retention the most. In other words, you want those inputs that yield a positive correlation. What you don't want is to spend time, money, and effort on inputs and have little to show for it.