The PRD was clear and reasonable. A real number, a plausible diagnosis, a proportionate fix. Design work was ready to start.
Improve discoverability of customer support.
Support discovery is cluttered, and more than 25% of NPS responders cite an inability to discover customer support.
Redesign the entry points to the care, support and help sections.
Two things about the brief bothered me. So before designing entry points, I wanted to watch someone actually fail to find support.
"Discovery is cluttered" and "redesign entry points" are the same sentence twice, which usually means the problem was inferred backwards from the fix rather than the other way round.
More than 25% selecting "could not find support" tells you what people ticked. It does not tell you what they were describing when they ticked it.
Seven students, all on long-term paid courses, in classes 9 and 10, all of whom had had their Student Account Manager changed after 45 days. I moderated all seven sessions myself.
Participants were minors, so the CX team contacted parents to obtain consent before any session was scheduled. One parent joined and participated in the call.
I asked what issue they had hit, how they tried to reach support, whether they knew what Vcare was and how to reach it, and if they had not used it, why not. Every session was logged against the same structure so the pattern would be visible rather than anecdotal.
Nobody had trouble finding support.
The stated problem did not exist. Support was discoverable and students knew exactly where it was. They were choosing not to use it.
One had contacted every available channel — SAM, email and Vcare — waited six months, and never got a resolution. He still reported that he could not find support. Another was satisfied with the support she received. She also reported that she could not find support.
The metric was real. Its label was wrong.
Students were not reporting a discovery failure. They were reporting that their problem had not been solved, and "could not find support" was the nearest box on the form.
Discoverability was never the problem. The obstacle was a mental model.
Every paid student has a named Student Account Manager. That is a person, assigned to them, with a relationship. Vcare is an institution. When something goes wrong, a fifteen-year-old calls the person they know, and falls back to the institution only when the person goes quiet.
That is not a navigation failure. It is a rational preference for a trusted human, and no amount of entry-point redesign moves it. We were about to spend a design and engineering cycle making something easier to find that people could already find and were deliberately not using.
I never disputed the 25%. That number was accurate. What I disputed was the inference drawn from it, and I disputed it with evidence about what respondents meant. Seven interviews cannot overturn a quantitative finding. They can, and did, establish that the finding was being read as something it did not say.
I wrote up the research and the recommendations. The PM took them to the stakeholders and the scope of the initiative was changed.
A single "Get Help" destination that becomes the fixed place students go, rather than a set of touchpoints scattered across the product.
With what each one is good for: chat, WhatsApp, email, phone, plus request-a-callback for paid students.
Published response times per channel, so a student choosing email knows it is hours and chat is immediate. If the underlying complaint is unresolved problems rather than undiscovered support, then managing expectation is closer to the real problem than improving findability.
And an open decision on whether the existing FAQ page should stay.
Request volume per channel, each channel's contribution share, and CSAT for issue resolution channel by channel. Note what these have in common — they measure whether problems get solved, not whether support gets found.
The SAM mental model, the deeper finding, went to the roadmap rather than into this phase. We fixed the channel experience first because it was tractable. Rebuilding how students relate to a named account manager is a service design problem, not a support-page problem.
The scope changed. Nothing shipped.
Vedantu moved into a period of losses and layoffs, and my engagement ended in April 2022 before the redirected work was built.
So the honest outcome is this: an initiative that was scoped, funded and about to start was stopped, and replaced with one aimed at the actual problem. The value delivered was a design and engineering cycle not spent on the wrong thing, plus a set of metrics that measure resolution instead of findability.
I would rather claim that accurately than dress it up as something that launched.
I reinterpreted an NPS item from the outside. The stronger move would have been to sit with whoever owned that survey and rewrite the question, so the instrument stopped producing the misleading signal at source. Fixing one initiative is worth less than fixing the input that will misdirect the next one.
Roadmapping it was the pragmatic call and I supported it. But 7 of 7 preferring a named human over an institutional channel is the most important thing the study found, and it deserved more than a placeholder.
"A number is not a finding."
"More than 25% said they could not find support" is a fact about what people clicked, and every survey instrument silently converts a range of experiences into whichever option is closest. When a brief arrives with a metric attached, the question is not whether the number is right. It is what respondents were actually describing when they produced it.