Evaluation of potential doesn't square with an album-based rating system very well. For example, when I rated the Sleigh Bells debut ahead of Janelle Monáe's The ArchAndroid as the best record of 2010 I was still looking forward to Monáe's followup more than Sleigh Bells', because I had a sense that Sleigh Bells had already maximized their potential and Monáe had room to grow. To put it another way, I think she's at least one hundred times more likely to make a five-star record someday. A more recent example: in 2012 I rated Gökçe's Kaktüs Çiçeği four stars, and Angel Haze's Reservation "only" three and a half, but I'm certain Haze's best future album will be better than Gökçe's.
Where does that sense come from, and what's it based on? Well, sometimes there are obvious, easily fixed problems with an early album: The arrangements and mix on Kaktüs Çiçeği are polished and precise, while the backing tracks on Reservation sound like they were whipped up in a weekend - Haze could make a much better record without improving a whit as a rapper or writer. Similarly, it can be a positive indicator when an artist's reach exceeds her grasp: Treats shows marvelous mastery of a very limited sonic palette; The ArchAndroid reaches into a wide range of genres and approaches but not always successfully, and it's easy to imagine Monáe digging deeper in plenty of different directions. (I try not to talk about Frank Ocean, but by these criteria he's a cinch to make a better album than channel ORANGE one of these days.)
Maybe this makes me an Auteur Theory goon, but I do cherish my suspicion that low floors imply high ceilings... In other words, any Jane or Joe can cut a bad record, but it takes a true original to get far enough away from stylistic norms to make a terrible one. For example, Lumpy Gravy is worse than anything ever put out by early peak/low ceiling icon Liz Phair. So rather than pegging artists by how good they can be on their best day, you might want to hear how bad they can be on their worst.