Since it's a pre-algebra class and you have the option of using KA full-time, I'm going to assume it's a middle school or high school class and thus that normal grade distributions aren't really necessary. In that case I would suggest simply splitting up grades based on completion/mastery of exercises, although I wouldn't recommend the full-time employment of Khan Academy for anything.

Hmmm.  Okay, I can tell you how my data on Khan related to end-of-grade testing for 10 fifth-grade classes.

So much of a student's final exam score is explained by how they did on Khan Academy over the course of the year.  Personally, I think that the KA mastery is a better indication of their achievement than a single three-hour test is . . .

One easy method you could use is assigning grade levels based on their percentage mastery in the Khan Academy skill.  A = 80% or higher, B = 60-80%, C = 40-60%, D = 20-40%, F = 0-20%.  That would have failed a ton of students in my situation -- but if you tell students up front that this is the grading scenario, then they'd know what to expect and might take progress in Khan more seriously.

Of course, this opens up the possibility of cheating (as does any grading system other than carefully-controlled final exam situations). There's a simple solution to that: in the final month of your class, introduce every student to a new Khan Academy account. Tell them that they may only sign into these accounts during class -- give them 20-30 minutes per day to work on the account. Tell them you can check to see if they sign in outside of class, and if they do so, you'll assume they're cheating and they'll fail. Their mastery score in their new account is their class grade.  Aaaaaaand you're done.  :-)