The prevailing orthodoxy is that the best way of assessing a teacher's performance, whether for appraisal or evaluation, is through observing him/her teaching. The idea is that the teacher provides a lesson plan according to state or district requirements for a pre-announced visit, often to a class and at a time chosen by that teacher. An impartial observer watches the lesson, and compares it both to the lesson plan and to some behavior-checklist which defines effective teaching. Post-lesson counseling then guides the teacher to some higher plane of practice. I disagree.
The observed lessons are rarely less than satisfactory, and frequently excellent or outstanding. By definition, all lessons are average so why are observed lessons so often so good?
The observation approach produces nothing more than smoke and mirrors or a dog-and-pony show. It reveals little if anything about the usual state of teaching and learning vis-a-vis that teacher. The teacher puts on a performance, and, particularly in the public sector where the teacher receives the observation-checklist in advance, performs to the observer's expectations. (Remember, what gets measured gets done.)
Even the students will play the game and perform appropriately, perhaps out of loyalty to "their" teacher.
Observed teachers craft a lesson-plan, perhaps the only such plan they produce that year. They research great plans and concomitant approaches, perhaps approaches they rarely if ever use. They spend hours decorating the room, arranging furniture, creating materials. Much about the lesson is unreal, if not actually false. Given the time and effort involved, multiple observations are unlikely and anything interesting produced can be seen only as a single data point.
So what do I advocate? First and foremost, look at student notebooks. They will tell what was received which is more important than what was sent. They will show you how well organized and planned lessons were. They will show you how much care, time and attention the teacher has given for example through teacher comments or grading. I or my division principals try to see two notebooks from every level every week. Records ensure all teachers and all classes are covered each semester, and particular staff or groups can be seen more often without being seen as threatening.
Secondly, ask the students. This can be done informally, perhaps when greeting or farewelling students at the front door, or when they come to collect their notebooks. This can also be done more formally, although that runs the risk of being seen as threatening.
Thirdly, I strongly endorse walk-throughs. As a school leader, I try to go into every classroom several times a week, for example when showing around prospective parents, and I ask my division principals to do the same. Each walk-through should provide no more than three data-points: what is on the board, what is on the board, Teacher Talking Time, student activities.
Fourthly, specific tests can reveal much and I do not mean standardised tests. If the entire ninth grade reads Of Mice and Men, have them all do the same paragraph test and grade them using the same rubric. If the entire fourth grade uses the same math book, write a test which requires students to show their methods as well as the final answer and grade them using the same rubric. Class and grade results will reveal much about teachers and their teaching.
However, above all, you must produce multiple data points over a period of time, also ideally produced by different observers. On its own, a single point does not show much. Taken together, they show patterns and trends and that is where an educational leader can take advantage of a teacher's strengths and address his/her weaknesses.
Get rid of lesson observations, do the above and see teaching and learning improve exponentially.