Julia Steiny describes the sort of data that school teachers can use to improve instruction:

Per the data-collection protocol, [Lonsdale Elementary School Principal Jeannine] Magliocco asks the kids at one table what they are learning today. As two girls speak over one another, we learn that this is a math class. They explain that while they need to be correct about the science they’re using to determine the space for each habitat, the lesson for right now, they emphasize, is about finding and plotting area and perimeter.
Magliocco scans her check list, finds “learning objectives are evident to the students,” and checks “evident.” The girls dive back into their work. I mention that the kids seem remarkably on task. Magliocco confirms that their teacher, Mike Maloof, is one of her most skilled.

The data collection process, though, doesn’t attempt to create a rigid, objective lever for evaluation. It does what must be done in an organic profession like education (perhaps any profession, ultimately) and creates a framework for subjective analysis of performance. People with knowledge of student-specific factors have to figure out where shortcomings exist and whether they represent failures, given the context in which they appear. (A classroom with significant extra-curricular problems might be doing very well even though another class performing at the same level would be doing poorly. Likewise a teacher with inadequate resources.)
In a final analysis, success will require a level of administrative authority and employee accountability that collective action and longevity — the claim that all measures must be objective and mechanically operable — just do not allow.