There's a numeric scale, but it's not going to make a lot of difference to folks being interviewed. When I host 'shadow' interviews - where we calibrate other Googlers on the scoring, the most important thing I ask is: Would you be OK if we hired this person onto your team and you had to be their coach -- and responsible for their quantity and quality of work, bugs and all -- for the next several months? If the answer is anything like "hell no", "no", "maybe", or "probably" then it's a firm DNH. If it's at "yes, but they're weak in area X", then put "maybe-hire, if-and-only-if someone else saw good strength in some other skill ". Moving up the scale, there's "Hire, with only minor reservations", and the ever-elusive "Definitely Hire!".

Each of those categories has finer sub-divisions on a point scale. A candidacy can survive a "maybe" or so, as long as there's at least one "definitely". We want to be sure that everybody considered for hire is excellent in at least one dimension. Five people saying "Meh, I guess so" is also DNH.

The issue is threefold for any tech hiring process:

• The opportunity cost of filling a spot with someone who's only-OK versus someone really excellent is huge.
• Four or five interviews at 45 minutes each is not really enough to get a full, rounded picture of anyone's total skill set, but that's all you get (at Google or anywhere else). 100% would be possible with a 4-week small project, but anything in between a day and a month doesn't buy much extra certainty.
• Depending on the size of your interview pipeline you have to pick a place to set your "probably, but" --> DNH bar. Google has a big pipeline, so we have the luxury of being able to set it pretty high.