Background
I launched a secured lending product on a large consumer fintech app in partnership with an NBFC partner, targeting a sizeable eligible user base within the app ecosystem. The product had strong top-of-funnel traffic driven by a large, existing user base — a significant audience, but one that wasn't converting into loan applications at any meaningful rate.
The Problem
Despite strong top-of-funnel traffic, overall funnel conversion — measured as leads created to leads submitted to the partner — trended at a very low single-digit percentage. A large majority of users were dropping off at the very first steps of the journey, before they had engaged meaningfully with the product.
The core issue wasn't awareness or reach. It was that a single, undifferentiated onboarding flow was being served to users with very different levels of product awareness, credit profiles, and intent signals. A first-time borrower and a highly creditworthy repeat user were seeing the same journey — and neither was being served well.
The Approach
I framed this as a persona problem before it was a product problem.
Step 1: Persona definition. I grouped users into distinct segments based on product awareness, credit profile, and creditworthiness signals — structured data that could be captured before the user even entered the active funnel.
Step 2: AI/ML decisioning layer. I fed these structured persona signals into an AI/ML model trained to drive product-journey decisioning at the entry point — personalizing the onboarding flow from the user's very first interaction, not after drop-off had already occurred.
Step 3: Two new onboarding flows. Based on the persona model outputs, I introduced two distinct onboarding flows — one designed for lower-awareness users needing more context and trust-building, and one designed for higher-intent, creditworthy users with a shorter, more direct path to loan submission.
Step 4: Growth and retargeting. I worked with the growth team to set up persona-based retargeting campaigns to re-engage dropped users efficiently and route high-intent customers to the partner with stronger conversion signals, reducing waste in the partner-side funnel as well.
Results
| Metric | Improvement |
|---|---|
| Overall funnel conversion (leads created → leads submitted) | ~3x improvement |
| Forward journey movement | ~50% improvement |
| Drop-offs at initial steps | ~35% reduction |
| Monthly disbursals | ~50% lift within one month of launch |
| Monthly disbursal volume | Scaled significantly within 3 quarters |
Key Takeaways
Persona before product. The single biggest lever wasn't a UI change or a campaign — it was understanding that the same journey cannot serve users at different stages of awareness and intent. Segmenting first, then building flows, is more effective than building one flow and optimizing it later.
Entry-point decisioning compounds. Placing the AI/ML model at the very start of the journey — before the user has invested time — meant the personalization had the highest possible impact on the metric that mattered most: early drop-off.
Partner-side conversion is a product problem too. Routing high-intent users with stronger signals didn't just improve our funnel — it improved the partner's conversion rate on submitted leads, which directly unlocked faster disbursal growth.