I didn't read the article, I don't think these kind has any value, but if you are here.. :)

i think it is rather binary, so

P_overall_success = 1 - (1 - P_success) ^ #attempts

and expected number of first success:

n = 1 / P_success

I don't see "cost" mentioned anywhere there.