In general, the PERT longest stretch and or ideal shortest stretch of dependent events should not contain k>=3 stacking risks <=0.50.

For example, one path with best probable outcomes:

[S]->[a]->[b]->[c]->[d]->[E]

a=0.73

b=0.49

c=0.33

d=0.50

By taking the 3 lowest/riskiest key deliverables in the best case outcome:

0.49x0.33x0.50 = ~0.08 likelihood of project success

This means one should restructure the project, redefine the goal scope, or shelve the project.

People that practice this approach say "no" a lot, but the firm does survive.

Best of luck, =)