Creating a "Checklist" of items that guarantee graduation will cause certain members of communities to aim toward achieving the checklist rather than focusing on site quality.
The emphasis should be less on numbers, and more on what defines a healthy community, with numbers as one indicator of whether a community is healthy.
"I know it when I see it" is a very effective tool for handling something as fuzzy as whether a community has the chops to make it.
Gaming the System
How many people review only to the point that they get a badge? How many people will post quick answers and then edit their posts to get the Enlightened badge or a bit more rep? How many people make minor suggested edits for the sole purpose of getting the +2 rep? How many people edited a community wiki post or made a bounty during winterbash just to get a temporary hat?
When you create incentive to game the system, people will game the system. When that's a meaningless hat or a badge, it's not so bad, but when the energy of a new community is focused on achieving arbitrary numbers rather than building the community, the community actually suffers because of it.
This is my #1 complaint about the Area 51 procedure -- the arbitrary cutoffs for getting to the next stage, along with the time limitations, encourage people to game the system just to push the process along to get to the actual community building. This encourages appeals to amateurs rather than experts in the sample questions and 'marketing' for the push toward private beta
The Emphasis Should be on Community Health
No matter how hard you try, numbers will never be able to define who is a good community member, they can only be used as indicators to keep an eye on to see where the community stands so that those community leaders can actually act on improving them.
The users that make a community successful are the ones who focus on the things that actually make a community successful. They define the scope in meta. They help out new users. They curate tags. They make helpful edits. They encourage community behavior. They create relationships with the regular users and make them feel like a part of the community.
Numbers should focus on letting the community know which areas they may want to take a look at and tackle as a community. For instance if meta is like a ghost town, that should be a red flag for the community to act on. On the flipside though, a lot of meta activity doesn't necessarily mean that the meta activity is good. If there are constant discussions of site design, or resolving conflicts between members, or rehashing the same arguments, that doesn't have the same value as meta posts that are defining scope or promoting community moderation. So how do you quantify "Successful Meta"?
"That's where the truth lies, right down here in the gut. Do you know you have more nerve endings in your gut than you have in your head? You can look it up. I know some of you are going to say "I did look it up, and that's not true." That's 'cause you looked it up in a book. Next time, look it up in your gut. I did. My gut tells me that's how our nervous system works."
- Stephen Colbert
Communities should be graduated when they feel right. They could be growing exponentially, have tremendous participation, but have an ideological rift within the community pulling the same site in two different directions. A community could be in beta for 4 years without seeing the growth that the SE team may expect, but show through their participation that the community is healthy enough to step out on its own two feet. There is no magical formula for what succeeds.
If anything, communities should be given more of an ability to speak directly with the SE Community Managers to discuss the "State of the Beta" to give SE a better view of how healthy the community actually is. The Area 51 stats and the regular site self-evaluations only provide numbers, if anything the focus should be more on the community members who are participating daily and what they feel about the health and strength of the community.