We see almost exactly the same thing people filling out the form see, except with a comment box at the top and a bunch of submit buttons representing different grades.
After doing this 16 times we have similar standards, so in practice we don't have to account for easy vs tough graders statistically. We do try to convince other partners to change their votes in borderline cases though. In the last phase we look at all the applications straddling the cutoff for the number of groups we can invite, and try to come closer to a consensus about what side of the cutoff each group should end up on.
No, he was probably telling the truth. The alumni do the first pass. But (a) they just vote no or maybe, (b) each application is seen by at least 3 alumni, and isn't filtered out till it has at least 2 more nos than maybes, and (c) we look at their overall stats to make sure no one is giving all nos.
pg don't you think the alumni would be biased towards startups/people they know. Or say no to those that could potentially become their competitors in the future.