FinTech · Financial Inclusion · Trust Design

Demand was never the problem. Confidence was.

Millions of small entrepreneurs in India are locked out of formal credit. P2P lending promised to reach them — but the platforms kept stalling, because neither side could trust what they couldn’t see. I designed a trust-first system that makes risk legible to lenders and lending legible to borrowers.

See the results View the work
45%
Borrower onboarding (baseline)
60%
Loan applications submitted
Tier 2–3
Underserved focus
Trust-first
Design principle
Scroll
lend.app/portfolio Dashboard OpportunitiesMy portfolioRisk advisorRepayments Portfolio health Invested₹2.4L Projected return12.8% At-risk loans2 On-time rate94% Repayments received Risk-scored opportunities A Tailor · Surat · ₹40klow risk · 11% · 12mo B Grocer · Pune · ₹65kmed risk · 14% · 18mo C Vendor · Nagpur · ₹30khigher risk · 17% · 9mo Browse all →

The lender dashboard — every opportunity risk-scored, portfolio health visible at a glance.

01 — The Trap

A market full of demand that the product couldn’t convert.

The borrowers were there. The lenders were there. The trust between them wasn’t.

India’s microlending ecosystem serves millions of small entrepreneurs shut out of traditional banking — no collateral, no formal credit history, no documentation. P2P platforms promised to reach them with faster capital and better returns. But for users in Tier 2 and Tier 3 cities, the experience collapsed under complexity and uncertainty, and the funnel leaked at exactly the moments that required confidence.

Here’s the thing leadership saw: fraud detection and regulatory compliance were already solid. The platform wasn’t broken on the back end. It was leaking on the front end — at every point a human had to make a high-stakes decision without enough understandable information to feel safe.
02 — The Insight

This wasn’t a lending problem. It was a confidence problem.

Research across lenders and borrowers, plus competitive teardowns of LendBox, LenDenClub and LiquiLoans, kept circling the same root cause. Both sides of the marketplace suffered from one identical thing.

The shared failure: a lack of reliable, understandable information at the moment of decision. Lenders couldn’t judge risk. Borrowers couldn’t understand their own repayment future. Same gap, opposite ends.

Lenders don’t need more data — they need confidence cues

This was the unlock. Lenders already drowned in partial dashboards and fragmented profiles. Piling on more raw data made decisions harder, not easier. What they needed was interpretation: a risk grade, a confidence level, a clear default signal — the data translated into a decision they could trust.

Borrowers need lending made legible, not simplified away

Underserved borrowers weren’t scared of loans. They were scared of rejection and of repayment terms they couldn’t see clearly. The answer wasn’t to hide complexity — it was to make the consequences of each choice visible and reversible before they committed.

Two users, one identical gap L The Risk-Aware Lender Individual investor chasing better returns “Am I making a smart decision with incomplete data?” FRUSTRATIONS — Incomplete borrower profiles — Unclear default signals — Manual, fragmented analysis Needs: confidence cues, not more data B The Underserved Borrower Small business owner, informal income “Will I be rejected, and can I actually afford to repay this?” FRUSTRATIONS — Complex applications — Fear of rejection — No visibility into repayment impact Needs: lending made legible, not hidden
FIG 01 The marketplace’s two personas, framed by the single gap they share: understandable information at decision time.
IN-DEPTH INTERVIEWS·BEHAVIORAL ANALYSIS·COMPETITIVE TEARDOWN·AFFINITY MAPPING·EMPATHY MAPS·USABILITY TESTING·AI RISK PROMPTING· IN-DEPTH INTERVIEWS·BEHAVIORAL ANALYSIS·COMPETITIVE TEARDOWN·AFFINITY MAPPING·EMPATHY MAPS·USABILITY TESTING·AI RISK PROMPTING·
03 — The Build

Trust, designed as a system — not bolted on as a feature.

One structural decision drove the whole architecture: separate the financial data from its risk interpretation. Raw numbers in one place, the meaning of those numbers in another. That separation is what let us add confidence cues without burying lenders in more dashboards.

1

Trust-first borrower profiles

Centralized borrower data with business context, income patterns and repayment behaviour — so a thin credit file becomes a readable story instead of a blank space.

2

Explainable AI risk signals

Alternative-data scoring, progressive credit grades and early default warnings — each shown with a confidence level and a plain-language reason, never a black-box number.

3

Loan advisor & simulator

Repayment scenario forecasting and a risk-return tradeoff visualiser. Both sides see the consequences of a choice before they commit to it.

4

Proactive monitoring

Portfolio health and repayment timelines surface risk early, turning lenders from anxious spectators into informed managers of their own exposure.

The structural decision: separate data from interpretation Card sorting with lenders revealed they conflated raw data with risk meaning. The IA pulls them apart. LAYER 1 · FINANCIAL DATA (THE FACTS) Borrower Profile Business context & income Financial Profile Repayment behaviour Orders & transaction history Raw, neutral, complete. No judgement — just what's true. AI LAYER 2 · RISK INTERPRETATION (THE MEANING) AI Risk Prediction Grade + confidence Loan Advisor Repayment simulator Early default warnings Interpreted, opinionated, actionable. The confidence cues live here.
FIG 02 The card sort exposed that lenders confused facts with meaning. The IA separates the two layers — data, then interpretation — which is what made confidence cues possible.
Borrower profile: from data dump to decision tool Each round moved closer to "confidence cue, not more data." LOW-FI · everything shown Problem: a wall of numbers Lenders still couldn't decide. MID-FI · grouped + grade data grouped A Grade added — but why "A"? Missing: reasons + confidence. HIGH-FI · explained grade A confidence stable income 12mo 7/7 on-time repayments informal income Result: a grade lenders trust.
FIG 03 The fidelity progression. The screen evolved from a data dump to an explained grade — the literal embodiment of "confidence cues, not more data."
Borrower profile — risk made legible Financial data on the left. What it means — interpreted — on the right. R Ramesh K. · Tailoring business Surat, Gujarat · 6 yrs operating · informal income Avg. monthly inflow₹52,000 Inflow consistency (12 mo)Stable Existing obligations₹8,000 / mo Prior repayments (platform)7 / 7 on time Income pattern · 12 months AI risk assessment A LOW RISK Model confidence 0.87 Suggested rate 11% · 12 months Why this grade Stable income across 12 months Perfect repayment history on platform Informal income — limited formal records ✓ No early default signals · monitored weekly Fund this loan Simulate repayment
FIG 04 The core screen. Raw financials on the left, the AI’s interpreted grade, confidence and reasoning on the right — the “confidence cue” thesis made concrete.
The trust system: every component reduces uncertainty RISK BADGE A B C An instant, glanceable verdict on each loan. CONFIDENCE METER 0.87 confidence How sure the model is — honesty about uncertainty. REPAYMENT TIMELINE Where repayment stands — past, present, upcoming. ALERT STATE ⚠ Payment 3 days late early warning · act now Risk surfaced early, while it’s still fixable.
FIG 05 The design system. Risk badges, confidence meters, repayment timelines and alert states — each one a confidence cue, not decoration.

The lender journey, before and after

BeforeAfter
Manual evaluation of partial profilesRisk-scored opportunities, ranked
High uncertainty at decision timeClear grade, confidence and reasoning
Defaults discovered lateProactive early-warning monitoring
Fragmented, anxious analysisOne legible view of portfolio health
04 — The Proof

The honest version of the numbers.

Straight talk: this was an early-stage engagement, so the strongest evidence is qualitative and directional, not a year of validated funding data. Usability landed at “OK, nearly good.” I’m showing where it fell short, because that gap is the most useful thing I can hand the next team.

What the work moved

Clearer
Onboarding and applications — complexity stripped out of the two steps where users stalled.
Higher
Lender confidence — risk visualization turned hesitation into informed decisions.
Earlier
Default detection — risk surfaced while it could still be acted on, not after the loss.

Targets the design is built to drive

↑ Funding
Higher conversion from application to funded loan
↓ Default
Lower default-risk exposure for lenders
↑ Retention
Stronger repeat-lender signals
Usability (SUS): solid, with a clear next step Scored 69.25 — “OK,” just under the 71 “Good” line. The gap is feature integration. 0 100 69.25 Poor (0–50) OK (51–70) Good (71–85) Excellent (86+) Strengths High learnability · low perceived complexity Gap → next iteration Cross-feature consistency · persona-specific flows
FIG 06 SUS landed at 69.25 — usable and learnable, with feature-integration consistency as the clear, honest next target.
The result I’d stake the project on isn’t a vanity metric — it’s that risk visualization measurably increased lender confidence and reduced decision friction. In a trust-first product, that is the conversion engine.
05 — The Call I Had to Make

Reach more borrowers, or protect lenders from defaults. You can’t fully maximize both.

I shifted the goal from “more loans” to “better loans.” Inclusion only works if lenders keep showing up.

This is the central tension of inclusive lending. Push inclusion too hard and you fund risky borrowers, defaults rise, and lenders flee — which ends inclusion for everyone. Tighten risk too hard and you rebuild the same exclusionary wall that traditional banks already are. The platform’s instinct was to chase loan volume. Research said that was the path to a default spiral.

So I made the call to design for better loans, not just more of them — using explainable risk signals and alternative data to bring thin-file borrowers in responsibly, rather than indiscriminately. That reframed inclusion as a sustainable system instead of a growth-at-all-costs bet. The win wasn’t maximizing one side; it was keeping both sides of the marketplace willing to participate.

What I’d do differently

Introduce financial-literacy tooling earlier, so borrowers build understanding before they hit the application, not during it. And invest in regional personalization sooner — a single flow can’t serve the linguistic and literacy range of Tier 2 and Tier 3 India, and the SUS gap hinted that one size was already straining.

The Role

Client — Early-stage P2P microlending startup
Role — UX Design Consultant: research, strategy, system design
Team — Satyajit Roy (UX Consultant) · a junior designer · a UI & interaction designer
Timeline — ~10 weeks: 1 mo research · 2 wk ideation · wireframes, design system, testing & final UI
What I owned — Research strategy, persona & mental-model definition, the risk-and-trust UX framework, AI prompting for risk-assessment workflows, and the reframe from “more loans” to “better loans.”