The year the funnel filled up — and the business slowed down. A root cause investigation.
Root cause: three junk affiliate partners flooded the funnel with unconvertible leads while the highest-converting channel (BTL) was simultaneously starved of volume.
Leads had grown 85.7% to 1.41M while enrollments fell 23.7% to 647 — the call-to-booking stage had quietly collapsed.
I ran a root-cause analysis across 8 BigQuery tables and 1.4M leads, testing six hypotheses before pinpointing the cause.
I proposed a recovery plan projected to add ~480 enrollments a year, anchored to rates already proven in 2025's own data.
Call coverage rose from 62% to 93.7% — the team was working harder than ever. The failure wasn't effort. It was what happened when those calls tried to become bookings.
93.7% of all leads were called. Coverage was never the problem — the quality of what entered the funnel was.
The divergence began in March 2026, the exact month affiliate volume surged. May 2026 set the all-time volume record — 584,819 leads — and produced just 231 enrollments. June 2025 had produced 640 from a third of that volume.
Every hypothesis was tested against BigQuery before making the report — quantitative validation first, qualitative signals second, then the fix.
Affiliate leads +58% (499K → 791K) while conversion crashed −89% (0.162% → 0.018%). Kollege sent 223,491 leads for 10 enrollments. Apna sent 10K for zero.
Three junk partners alone flooded 371K leads into the funnel — consuming ~348K rep calls that could have gone to leads capable of converting.
BTL — walk-ins, school visits, events — held the highest conversion of any source (0.178%), yet its volume was cut 69% (428K → 131K).
Worse: BTL leads waited an average 40.7 days for a first call — the longest delay of any source. High-intent walk-ins had already enrolled elsewhere.
22% of all call objections were product mismatch — Wrong Programme (9.4%), NEET (5.3%), Nursing (3.3%), Wrong Inquiry (3.9%). These leads could never enroll.
Campaign creatives were never updated after the portfolio changed — discontinued programmes like Radiology were still generating inbound leads.
Leads called within 24 hours convert at 0.282%. At 8–14 days: 0.082% — a 71% penalty. PM leads averaged 8 days to first call.
The best-converting segment — PM leads in Mumbai / Bangalore / Delhi — waited 7.3 days on average, despite the highest conversion potential in the base.
Uni Pass payment rate: 1.8% at April launch → 54.9% in May (working) → 5.3% in June. 2,698 of 2,849 June bookings didn't pay.
May proves the gate can work — June's collapse was operational, not structural. Manual overrides with no formal criteria made the quality filter unpredictable.
All five funnel stages degraded simultaneously — a shared structural cause. Rep performance gap widened from 65.8 to 78.6 points; bottom quartile fell from 21.4% to 3.1%.
Portfolio went from 2 to 4 programmes across 2 sectors — but bot scripts, affiliate briefs, campaign creatives, and rep training never caught up. Parent/Family Approval grew 4,733% as a drop reason.
Speed-to-lead decay — conversion rate by time to first call:
* 30+ day elevation = walk-in and referral revival, not cold digital leads. Calling within 24 hrs converts 3.4× better than after 8 days.
of all conversions happen within the first 10 calls. Duration falls from 211 sec (calls 1–5) to 61 sec (30+) — early call quality beats late-stage persistence.
combined conversion of the three junk partners — Kollege, Apna, Careers360 — across 371K leads. The direct affiliate source converts at 2.176%: a 360× spread.
offline demo show-up vs 6.7% online — 1.6× more effective. And 83% of converted leads enrolled without a formal demo at all, via walk-in or campus visit.
growth in Parent/Family Approval as a post-demo drop reason (3 → 145 cases). The follow-up process talked to students — never to the decision-maker.
zombie leads — not enrolled, not lost, no activity 60+ days. At the 30-day band rate, a zero-cost bot re-engagement is worth 15–30 recovered enrollments.
of the entire database — 946,653 leads — missing a city tag. No routing, no city SLA, no geographic analysis. Fixable for free at the form level.
Every target below is a rate already achieved in 2025 or observable in the current dataset. Each one is traceable to a specific data point.
| Metric | 2026 Current | Target | Basis | Est. Impact |
|---|---|---|---|---|
| BTL conversion | 0.178% | 0.230% | 2025's own actual rate — needs same-day SLA replacing the 40.7-day average | +68 enrollments/yr |
| Call-to-booking | 0.64% | 1.50%+ | Remove 371K junk leads — reps handle fewer, better leads | +~100 enrollments/yr |
| PM response time | 0.054% | 0.150%+ | 24-hr called leads convert at 0.282% in our own data — target set halfway | +260 enrollments/yr |
| Overall conversion | 0.046% | 0.080%+ | Combined effect — affiliate cleanup + speed fix + BTL restoration | +480 enrollments/yr |
| Zombie re-engagement | 0 (inactive) | 15–30 | 15,666 dormant leads × 30-day band rate × 50% re-engagement | One-time recovery |
Conservative by design — these assume partial implementation and zero improvement in the 22% mismatch problem. Fixing that adds another 30–40%.
Enrollments fell at the bottom of the funnel, but the cause sat at the very top — lead sourcing — and in a cross-cutting change (the portfolio expansion) that no single funnel stage owned. All five stages degrading simultaneously was the tell.
Recovery estimates built on 2025's own achieved rates and observable in-dataset behaviour are credible in a leadership room. Projections built on optimism are not. The whole report was designed so every claim is traceable to a query.
29% of leads missing city tags meant a third of the base couldn't be routed, SLA'd, or analysed geographically. The cheapest fix in the entire report — one mandatory form field — improves the accuracy of every future analysis.