Cognitive bandwidth is the constraint —
not awareness.
New teachers don't observe less because they don't care about what students are doing. They observe less because delivery consumes most of their cognitive bandwidth. Managing the explanation, staying on time, handling transitions, tracking the board — all of these compete for attention with the observation of 25 students simultaneously. Observation loses, because it is not automatic yet.
Expert teachers observe more because they have automated the delivery. The explanation runs on procedural memory; the transitions are practised. This frees cognitive capacity for observation, which then runs alongside delivery rather than competing with it. The implication is that observation cannot be fully developed in the first year of teaching — and that practices which accelerate the automation of delivery (shared lesson plans, stable routines, AI generation) indirectly accelerate the development of observation skills.
For teachers at any stage, the three practices in this article partially compensate for the bandwidth constraint by making observation more efficient: you observe more with less cognitive effort because the practice is structured rather than open-ended.
Where you stand determines
what you can see.
Most teachers spend most of their time at the front of the room — near the board, near their desk, visible to students. This positioning is comfortable and allows easy reference to materials. It is poor for observation: it creates a consistent angle that makes the back rows hard to read and places the teacher in a position where students perform for them rather than work independently.
Move to the back of the room during independent work. Walk along the back wall and look forward — you can now see every student's work from above and behind. Students cannot turn to face you for approval. The angle reveals which students are on-task, which have stopped, and which have the systematic wrong answer without you standing over any individual.
Spend 2 minutes at the start of term identifying them: the corner behind the door, the far-left row if you stand at the right, the student whose desk faces away from your usual circulation path. These are the students most likely to be off-task without you noticing. The awareness alone changes your positioning enough to reduce the effect.
During whole-class instruction, stand slightly to the side of the board rather than directly in front of it. This allows you to face both the board and the class simultaneously. During group work, stand at the edge of the room — not inside a group — so you can see multiple groups at once without being absorbed into any one conversation.
Observe the same students in the same order —
every lesson.
Random observation is inefficient. You repeatedly check the students who are visibly on-task (because they are easy to confirm) and miss the students who are quietly off-task or confused. A planned scanning sequence checks every student in a predictable order — not to catch misbehaviour, but to maintain a consistent picture of the class's work state.
The scanning sequence is simple: divide the class into four quadrants (front-left, front-right, back-left, back-right). Scan one quadrant per pass during the proximity walk. In a 20-minute independent work period, you will have scanned every student at least twice. The second scan tells you whether the picture has changed — which is the data source for micro-pivot decisions.
One thing to watch for.
Every lesson.
Open-ended observation produces open-ended data. “Watch the class” produces a general impression. One specific focus question produces a specific answer — which is what A2's pivot decision framework requires to operate at its most precise.
If yesterday's exit ticket showed that most students got Q2 wrong by applying [specific misconception], today's observation focus is: 'At what point in today's explanation do I see that misconception appear in students' written work?' That specific focus produces the specific data that tells you whether today's revised explanation resolved the issue.
At the 15-minute mark of the lesson, briefly scan specifically for the thing you set as your focus. 2 minutes. What do you see? If the misconception is gone, the revised explanation worked and you can note it in the iteration log. If it's still present, you have real-time confirmation that a mid-lesson stop is needed.
At the end of the lesson, your observation focus question and what you found becomes Input 2 of tonight's AI regeneration prompt (C4/A3): 'I was watching specifically for [focus]. I observed [finding].' This is the highest-quality observation note you can write — specific, focused, and directly tied to yesterday's diagnostic data.
C2 leads directly into
C3 and C4.
The real-time observation practices in C2 produce the in-lesson data that feeds C3's exit ticket design (the pre-lesson focus tells you what the exit ticket needs to check) and C4's iteration workflow (the observation note tells the AI what to fix). All three clusters operate as a single data-gathering-and-responding system when the practices are integrated.