Some subscribers to Math-Teach might be interested in a recent post "The Value of Student Evaluations of Teaching" [Hake (2012)]. The abstract reads:

*******************************************
ABSTRACT: Bill Brescia of the DrEd list wrote: "We are looking for best practices to solicit student feedback on the teaching quality of individual faculty members where in some cases a fairly large number of faculty are involved in a single course. What mechanism do you use to survey the students?"

The value of student evaluations is a hotly contested topic, witness the 496,000 hits at <http://bit.ly/TqEilE> generated by a Google search for "Student Evaluations" (with the quotes) on 26 Nov 2012 12:58-0800.

Judging from my own experience, and after careful consideration of the above hits, my opinion is that:

a.  Student Evaluations of Teaching (SET's) are useful for gauging the *affective* impact of teaching, but are worse than useless for gauging the *cognitive* impact - see e.g. "Student Evaluations of Teaching Are Not Valid Gauges of Teaching Performance - Yet Again!" [Hake (2012)] at <http://bit.ly/KGK687>

b. The cognitive impact of teaching is best measured by average normalized pre-to-post-test gains on "Concept Inventories" <http://en.wikipedia.org/wiki/Concept_inventory> - see e.g., "The Impact of Concept Inventories On Physics Education and It's Relevance For Engineering Education" [Hake (2011)] at <http://bit.ly/nmPY8F>.
*******************************************