One way to run achievements is to encourage breadth. Have achievements for doing everything once or a few times. Complete this dungeon, kill ten rats, cook every type of pie. I advocate this for Explorers, of which I am one. It encourages players to see all of your content, therefore extending the time until they get to the end and complain that there is nothing to do, while rewarding long-duration playthroughs. It can increase retention that way while serving as a checklist for players looking for something to do. It can also be expensive, because you need an entirely new thing every time you add achievements.
One way to run achievements is to encourage grinding. Have achievements for doing things repeatedly. Complete this dungeon 10 times, kill ten thousand rats, cook one hundred pies. I advocate this for Achievers, which I am secondarily. If you are going to track and display how many rats Bob has killed, as I think you should, you can then tack awards onto it. This is easy to develop, because you just add a name for each power of 10 for each enemy type. You get that Explorer award for the first kill of each enemy type, then add another tier for 10, 100, 1000, …, 100,000,000. You do not even need names for all of them at the start, because you will have lots of time between the time he becomes Bob the Super-Epic Ratslayer for 100,000,000 rats and Bob the [prefix] Ratslayer for 1,000,000,000 rats. Then add meta-achievements, for having killed 100 of everything. This means that there is always something more to do, more to Achieve, and it can increase retention with a never-ending checklist. Not everyone will pursue it, but if people will pay you to run on an achievement treadmill, set it up and take their credit card information.
The latter idea came to me first, and I find myself turning away from it. First, I would not want to do it myself. If I played a game for 10 years, it might be nice to know that I had killed 2,405,353 goblins, but then again I might see that (or my /played) and think of what else I could be doing with my life, and I certainly do not need the logarythmically extending bars ever before me. Also, I think it would encourage aberrant gameplay. If there is some reward, any reward, for killing ten million rats, someone will do it. You may not like what they do to optimize rat-killing time. Without the impetus to exploit them, certain holes and bugs could just sit there until you get that far down your priority list, instead of having the forums burning up with the implications. You can also burn out players that way, ones who start to see through the grind and wonder why they are killing that next million rats.
I suppose that is a financial decision. Which is the larger pool of subscriber dollars: retention of obsessives or loss of burnouts? Since I personally would not want to be racing that treadmill, I would fall on the “loss” side. But you will always make more money on the mass market betting against my preferences, so let the eternal achievements go forth for greater profits.