Does the AI coordinator help?
The coordinator steps in when a student's plan needs to change. To know whether that really helps, it is tested fairly: after a mock well below the plan or a missed checkpoint, 20% of cases are picked at random and left to the normal plan with no extra change. The two groups are then compared on the same measures, fixed in writing before any data existed. Every result is shown here, including the ones that are not good news. Updated 2026-11-16.
- Coordinator stepped in
- 0
- Left to the normal plan (comparison)
- 0
- Always helped (student asked, missed days, burnout)
- 12
- Waiting for their 14 days to finish
- 12
After a full mock well below the plan
Accuracy on the weak chapters (14 days after vs 28 days before)(main measure)
Collecting data- Coordinator
- – (0)
- Normal plan
- – (0)
- Difference
- –
Next full mock vs the mock that triggered it
Collecting data- Coordinator
- – (0)
- Normal plan
- – (0)
- Difference
- –
Daily missions done in the next 7 days
Collecting data- Coordinator
- – (0)
- Normal plan
- – (0)
- Difference
- –
After a missed checkpoint
Accuracy on the weak chapters (14 days after vs 28 days before)(main measure)
Collecting data- Coordinator
- – (0)
- Normal plan
- – (0)
- Difference
- –
Next full mock vs the mock that triggered it
Collecting data- Coordinator
- – (0)
- Normal plan
- – (0)
- Difference
- –
Daily missions done in the next 7 days
Collecting data- Coordinator
- – (0)
- Normal plan
- – (0)
- Difference
- –
Changes that are never held back
When a student asks for a change, misses days or shows signs of burnout, the coordinator always helps, so there is no comparison group. These are plain averages after the change; they cannot show cause.
- After a burnout warning: Daily missions done in the next 7 days 0 (1)
- After missed days: Daily missions done in the next 7 days 0 (4)
The rules
A verdict needs at least 30 cases in each group. "Helps" means the whole 90% range is above zero; "hurts" means it is below zero. The range treats all of one student's cases as one student's evidence. If the main measure ever says "hurts", the coordinator stops that change for everyone automatically.
The comparison group still gets the full normal plan: daily missions on weak chapters, revision and mocks. No student data is shown here, only counts and averages.
What this cannot show yet: small effects, and the effect on the real NEET score, which needs results after the exam.