Are there methods for determining the exploitability of bugs, after a crash is found in fuzzing?

I have found this, http://ift.tt/1OgR9De

but it says that only "a full root-cause analysis can ensure that you have correctly diagnosed whether a given crash is exploitable". Is it possible to do such analysis without access to source code ? How ? If not, what´s the closest to "full root cause analysis" that can be done without source code ?

Thanks,

Rafael