There is no universal barrier that would be considered "Good Sales" that would cover most games; as an example 1 Million units sold would (probably) be considered very successful for Boom Blox, and the same sales would make Grand Theft Auto 4 one of the biggest flops of all time. I do think that the following schedule is (probably) a pretty good starting point though:

I would say if it meets expectations, it's good sales.

If it exceeds greatly, awesome sales.

underperforms, poor sales.

Example: if GTA 4 sold 2 million first week, people would consider it terrible sales. If Wiifit does it? awesome sales.

depends on the region

depends on the cost to get the game to market

depends on the title

i think that over 400,000 is considered as a good game!!

500,000 is gold, 1 000 000 is platinum......

It is totally dependant on each game.

I would say there are two measures of sales, one is the specific titles expectations (from the public, us or journalists/analysts) the other is how profitable it becomes due to good sales.

For instance if GTAIV on both platforms failed to make 12 million sales it wouldn't be considered good sales at all when previous versions have all sold so many more copies (double it with San Andreas)

But profit wise Take2 may have broken even already, assuming $10-20 per copy goes back to Take 2 they needed to sell roughly 6 million (@$20) or 12 million (@$10)

That is assuming the $100 million development figure is correct, and adding $20 million for advertising etc.

That of course doesn't factor in Microsofts money either, but as I am not sure on the specifics of that I ignored it, I have heard it was a loan to Rockstar that is to be repayed, or perhaps it was purely money to Rockstar and might not have made it to Take 2s hands (although I think R* is owned by T2 so I doubt this)
---------

That is an example of a massively high budget game though, for which profitability needs incredible sales, MGS4 for instance I am not sure will ever make Konami a profit because it will surely need 5 million sales if not 6 or 7 just to break even.

 I like your table and think it is a good place to start. After that, you can factor in publisher expectations, and other factors, like if it was a port or not.

Wiifit is probably not a good example.  It will be supply constrained as it has a fairly expensive hardware component, unlike a regular game that might be $1 or $2 at most...

Some shovelware is probably happy with 100,000 unit sales.

What's the source of these numbers?

The easiest way to determine good sales is how quickly they announce the sequel. Succesful games go on and on while flops are never heard from again.