Customer Discovery for a Credit-Building Product

Role: Research ops & voice/chat tooling · Phase 1 guides + analysis · Phase 2 interviews + analysis

Duration: Two phases, ~2 months

Client: Pre-launch fintech startup (India)

Agency: 1990 Research Labs / Canvass

Research methods: Structured voice interviews (110) + in-depth video interviews with screen-share exercises (card sorting, ad testing)

Quick snapshot

• Project: Customer discovery for a pre-launch fintech product that turns routine utility-bill payments into CIBIL credit-building behaviour.
• The question: Who should the product launch to first, what would make them trust it, and what would keep them around long enough to see their score move?
• My role: I designed the Phase 1 interview guides, built the recruitment tooling, and owned the analysis across both phases , plus the final 10 in-depth interviews myself. (The 110 Phase 1 calls were fielded by my team.)
• Team: Agency research team (1990 Research Labs / Canvass).
• Constraint: Pre-launch product, no existing users , every participant recruited cold, and trust had to be evaluated before anyone had touched the product.

Impact

The research was delivered as a go-to-market strategy for the launch - mapped across the full customer journey, from first touch to referral, and spanning launch audience, messaging and trust, onboarding, retention, and referral. (The specific recommendations are the client's and stay confidential; what follows is the research that shaped them.)

Its biggest effect was reframing how the team understood the launch:

My role and how I analyzed the data

• Phase 1 : designed and analyzed. I wrote the interview guides used across all 110 voice interviews (which my team conducted), then coded and synthesized the results.

• Recruitment infrastructure : built end to end. The full Meta → WhatsApp → AI voice-agent chain, including chatbot/tooling evaluation and campaign setup.
• Phase 2 : led the deep dive. Conducted the final 10 in-depth interviews myself and led the analysis.
• Helped pivot to recruitment-agency sourcing and human-led calls when the automated funnel didn't fit cold-ad respondents, and partnered on the transcript-to-searchable-Q&A flow used for analysis.

For synthesis, I mapped each Phase 2 interview against the highest-risk pre-launch decisions, then triangulated those patterns with my Phase 1 coded moments to separate one-off reactions from repeated behavioural signals.

Does the problem actually exist?

Yes , and the demand was already there. When the concept was explained, 79% showed positive intent; most participants needed credit for a concrete goal (a home, a vehicle, a business), and about half had already tried an alternative like FD-backed credit cards. The product was meeting a problem people were already trying to solve. Credit rejection was costing them real, time-bound opportunities, and they were already hunting for a way out.

"I've been paying electricity bills for years. If I'd known this could build my credit score, it wouldn't have even felt like effort - and my wedding loan would have been so easy to get." — Participant, CIBIL Aware – Rejected

"It's a chicken-and-egg problem. Without a credit card you can't get a score, but without a score you can't get a credit card. So I went looking for alternatives." — Participant, CIBIL Aware – Never Taken a Loan

Key insight: The problem isn't financial irresponsibility, it's institutional invisibility. These users understand credit; the system just doesn't recognise them.

The audience: not personas, but stages of one journey

The research showed users weren't fixed types but people at different points on the same credit-rejection path:

• CIBIL Aware – Rejected : recently rejected, understands the problem, actively trying to build credit.
• CIBIL Aware – Never Taken a Loan : knows their score is zero, looking for solutions but not yet urgently.
• CIBIL Unaware : rejected, but hasn't yet connected the rejection to their score.

"People had really pumped me up. But as soon as my loan got rejected, I came crashing down." — Participant, CIBIL Aware – Rejected

"I feel, did I commit any crime? Everywhere I go, whatever I do, they take so much inquiry about me." — Participant, CIBIL Unaware

Key insight: Motivation, emotional urgency, and active solution-seeking all peaked in the recently rejected stage, which is what made it the clearest starting point for launch.

The trust gap: they want to trust you, but the internet has burned them

With no product to try yet, participants raised their guard unprompted.

"AI-generated - no, not for a financial app. If this had been a makeup or a jewellery brand, I'd have trusted it. But this is a finance brand." — Participant, CIBIL Aware – Rejected

"If I've allowed them to take my details - my family photos, my contacts - they can steal my details. Anything can happen." — Participant, CIBIL Unaware

What did build trust was proof, not polish:

"I'll search this app in the RBI guidelines list. Whether the government considers you trusted also matters." — Participant, CIBIL Unaware

Key insight: For this audience, trust is a research-before-trust behaviour, not a feeling, which changes what a financial product has to prove, and how it proves it.

A mental-model finding: "is this just another payments app?"

One recurring signal from the interviews: when the product was framed around bill payments, participants instinctively compared it to apps like GPay and PhonePe, and judged it on those apps' terms (free, instant, cashback).

"I'm only making bill payments, and this is becoming free through other apps." — Participant, CIBIL Aware – Never Taken a Loan

Why it mattered: left unaddressed, the product would be judged in the wrong category, measured against free UPI apps on terms it was never meant to compete on.

The research ops pivot (and what it taught me)

Phase 1 was designed as a self-serve funnel: Meta ad → WhatsApp chatbot → scheduled AI voice-agent call. It broke in two places:

• Chatbot fit: respondents replied in free text; the bot expected structured selections.
• Interview expectation: cold-ad respondents wanted product information, not an AI-led research interview.

We pivoted to recruitment-agency sourcing and human-led calls - that version produced the 110 Phase 1 interviews.

Limitations

• Phase 1 was self-selected through ad response, so it skews toward people already engaged with the credit-building problem.
• No rural representation: findings speak mainly to urban and semi-urban credit-rejection experiences.
• Phase 2 was deep, not broad: directional rather than statistically representative.
• All reactions were to a pre-launch concept and mockups, not a live product.