Hear about a weak teacher before the parents do
A monthly student-feedback loop cut the time to act on a teacher problem from three or four months to about one.
Time to act on a weak teacher3–4 months to About 1 month
The situation
Teacher problems surfaced only when parents complained, usually three or four months too late. With 800 students, reading thousands of feedback forms by hand was impossible.
What I did
Every month, students across all centres rate their teachers on a few points, such as syllabus coverage, clarity, doubt resolution and punctuality, and answer three or four open questions in their own words. AI reads the free text, groups it into patterns and flags batches likely to cause trouble in the next one or two months. A person checks every flag against the actual comments before anything happens.
A flag starts a conversation, then a month of observation, then support or retraining, then a decision.
What changed
| Measure | Before | Now |
|---|---|---|
| Time from first signal to a decision on a teacher | 3–4 months | About 1 month |
| Monthly review effortOne trained person | 5–6 days by hand | Under 2 hours |
| Teachers retrained from the feedback | None attempted | Two, in Maths and Chemistry, now among the best reviewed |
| Student candourBecause feedback leads somewhere | Hesitant | More honest feedback |
What it cost, and where it stops working
The AI flags a pattern. It does not decide. So far the batches it flagged did turn out to be problem batches, but that is a small sample. I treat a flag as a prompt for a conversation, never a verdict.
What it shows
Teacher development as an operating routine, not an event.
How this was measured
Internal monthly feedback records. Decision time runs from the first feedback signal to a decision on retraining or replacement. Review effort compares hand analysis with the trained process. Unaudited.
Kind of problem
Teacher quality
Related