Case Study 02
Six months into building State PSC, we ran our first proper NPS survey. It came back at 23, against an EdTech benchmark of 60+. Six months later it was 67, with three times as many students answering.
First Survey
23
NPS
45% promoters · 22% detractors · 725 responses
Six Months Later
67
NPS
76% promoters · 9% detractors · 2,175 responses
Participation
3×
More students answering
725 → 2,175 responses
01 · The Problem
For its first six months, State PSC ran on batch ratings, educator ratings and attendance. Useful signals, but none of them asked whether students would recommend us.
The first NPS survey answered that: 23. For a business that grows on referrals, upgrades and word of mouth in small exam communities, that number put the whole growth plan at risk.
02 · Diagnose
We called all 160 detractors. My category managers ran the calls, a proper interview each time, and I went through every one of them. Alongside that, I compared notes with other PW category leaders, studied Study IQ and Drishti IAS, and split the survey into the five parts of the product a student actually uses.
| Part Of The Product | Unhappy Before | After |
|---|---|---|
| Contact Us (support) | 39% | 15% |
| Test quality | 34% | 10% |
| Doubt support | 32% | 12% |
| Faculty | 28% | 8% |
| Practice papers (DPP & DMP) | 23% | 7% |
“Unhappy” means students who rated that part 3 or below. Every one of the five was a product and operations problem, not a marketing one, so every fix went into the product.
03 · Fix
Newspaper subscriptions for the content team (The Hindu, Indian Express, Varshiki), so current-affairs questions shipped the next day. Question setters hired from people who had sat the exam in the last 12 months, so papers matched the latest format.
34% → 10%
Students unhappy with test quality
The faculty bench was strong on Prelims and thin on Mains. Hired three General Studies specialists to cover Mains in depth.
28% → 8%
Students unhappy with faculty
Made it standard for every educator to keep the last 15 minutes of each 1.5–2 hour lecture for live doubts. Inbound doubt tickets fell 40%, and the rest got answered faster.
60% → 87%
Doubts resolved within 24 hours, in one month
Replaced a fixed five questions a day with weightage-based sets: 10–15 questions on high-weight topics, 3–5 on low-weight ones. More practice where it counts, at no extra cost.
23% → 7%
Students unhappy with practice papers
Launched a category-owned WhatsApp support line, with the category manager triaging each query straight to the right faculty, ops or tech owner, instead of a slow central queue.
150–200 → 50–60
Support queries a day within a month, because issues got fixed faster
04 · Outcome
Promoters rose from 45% to 76% and detractors fell from 22% to 9%, taking NPS from 23 to 67. Responses grew from 725 to 2,175, so the higher score came from a much wider, more honest sample.
05 · Takeaway
NPS moved because the product got better, not because the survey did. Every fix traced back to a detractor call. Today I’d run that same loop, detractor calls, ticket triage and content freshness, with AI agents doing the legwork.