A test-prep coaching institute launched an LMS for 8 000 students in 11 weeks
The institute in one line
A test-prep coaching institute (JEE + NEET) with 5 centres across one metro region, 8 200 students enrolled at a time, 260 faculty + staff. Hybrid model: classroom + online lectures live-streamed to satellite centres.
The before
Three problems, all human-time:
- Faculty wrote quizzes by hand. Every week, every faculty across every batch wrote 1-2 quizzes per topic. Across 260 faculty, that was thousands of hours of question-writing.
- No central LMS. Each centre ran their own Google-Drive folder. Students at the smaller centres got less material because their faculty had less time to upload.
- Lead-to-enrolment was a CRM black hole. Walk-ins, web enquiries, referrals, and demo-class attendees all sat in different spreadsheets. Conversion-rate per source was unmeasurable.
What we changed
In 11 weeks, with two-week sprints:
Sprint 1-2 — LMS foundation. Course tree, lesson uploads, video hosting, mobile downloads. Faculty migrated existing material into the central LMS over 3 weeks.
Sprint 3-4 — Batch-based fee plans + lead funnel. Enquiry capture from the website + walk-ins, status pipeline (new → contacted → demo-taken → fee-quoted → enrolled). Per-source conversion attribution.
Sprint 5-6 — AI Quiz Generator rollout. Faculty upload a lesson PDF, click "generate 10 questions". The system produces a mix of MCQ + short answer + descriptive. Faculty reviews each, edits or discards, then publishes. Average time per quiz dropped from 35 minutes to 5.
The after
| Metric | Before | After |
|---|---|---|
| Faculty time per quiz | 35 min | 5 min |
| Quizzes published per week (institute-wide) | ~60 | ~340 |
| Students with > 1 attempted quiz/week | 22% | 78% |
| Lead conversion-rate (web enquiries) | unmeasurable | 8.4% → 11.2% (after 3 months of A/B testing call scripts using the funnel) |
| Time-to-respond to web enquiry (median) | 8 hours | 22 minutes |
The AI Quiz Generator number is the one the institute's owner refused to believe until the third faculty member showed her in person. She'd budgeted for hiring 4 more faculty in 2026 just to keep up with content load — she now doesn't need to.
"We had AI everywhere in our marketing. Now we have AI in our operations. The students see the difference."
— Director of Academics
What we'd do differently
Two things:
One: we let faculty self-onboard onto the LMS without a structured training plan. The 30% of faculty who were comfortable with software figured it out; the other 70% needed hand-holding. Train-the-trainer in week 1, then peer-onboarding from week 2, not pure self-serve.
Two: we shipped the AI Quiz Generator in week 5 (early). It worked brilliantly but became the single most-used feature within 2 weeks, which drove our AI cost forecast 4x higher than planned. We added the per-feature quota guard (already on the roadmap, but accelerated by this) — but if we'd had it ready on day 1 of the AI rollout, the rollout would have been smoother.
Now, every AI feature ships with quota + kill switch as launch-day, not later.
Run a coaching institute? Talk to us — we'll show you the AI quiz generator on your subject content live.