PROBLEM
People started loan applications and didn’t finish them.
- Fundu · Consumer lending · 2019–22
- 8 min read · 11 chapters
Designing a lending app people came back to.
A consumer loan aggregator for India, built from zero as the only designer. Live in five months, then iterated for three years.
- ROLE
- Sole UX/UI designer, reporting to the CEO
- TEAM
- CEO as PM · 15–20 engineers · marketing owned acquisition
- WHAT I DID
- Research, IA, flows, UI, handoff · Android · 100+ screens
- TIMELINE
- Jul 2019 – 2022 · live Dec 2019
Play the problem
scrub 27 screens down to 10
- 74%
- of applicants finished, after the redesign
- ~20%
- fewer drop-offs
- 27 → ~10
- onboarding screens
- 10
- product surfaces, designed solo
74% is an absolute completion rate, not a lift. 3M+ installs in year one were marketing-led, so they aren’t claimed as a design outcome.
The case in 30 seconds
busy? this is the whole story
INSIGHT
Ask once, ask late. Collect the shared core one time, defer the rest.
WHAT I BUILT
A ~10-screen flow, then a 10-surface ecosystem to come back to.
IMPACT
74% finished the application. Drop-offs fell by about a fifth.
“Compete with a market leader.”
That was the brief. No target, no defined user, no success metric. It started as a side project.
- No target user
- No success metric
- No defined problem
defining the problem was the first deliverable
01 · The brief
The research came from live campaign data.
The parent company ran lead-generation campaigns for the competitors. That meant real performance data on the Indian lending market before a single screen existed.
Sat with marketing and analysed campaign performance across competitor accounts.
Ran hands-on teardowns of the competitor apps myself.
Looked at which messages converted, and for whom.
Two cohorts, split on financial literacy × ticket size.
Suresh, 35
Higher literacy · larger ticket
- CONVERTED ON
- Rate comparison, speed
- NEEDED
- Options and control
Priya, 28
Lower literacy · smaller ticket
- CONVERTED ON
- Eligibility, reassurance
- NEEDED
- Guidance and clarity
personas came out of the evidence, not before it
02 · Research
People started applications. They didn’t finish them.
27 screens, one input per page
Built as a questionnaire, not a form.
Every lender asked again
10+ lenders, each with different requirements. Apply to three, fill three sets of details.
The heaviest asks came first
Documents and verification landed before the user had seen a single offer.
one question per screen, 27 times over
03 · The application problem
Ask once. Ask late.
One common data core, collected a single time, then mapped to every lender’s requirements behind the scenes.
Common core
The shared set every lender needs, captured once at onboarding.
No repetition
Apply to any lender without refilling anything.
Deferred documents
Lender-specific uploads only after the user picks a lender.
Motivated ask
By then they’ve seen an approved offer and a real number.
Sequenced by motivation
- Common coreasked once
- Offers shownreal numbers
- Picks a lendercommitment
- Lender documentsasked late
the heaviest ask sits at the point of highest commitment
27 screens → ~10. Same data, a fraction of the taps.
BEFORE · 27 SCREENS
AFTER · ~10 SCREENS
27 screens10 screens
27 screens.
The new flow, screen by screen
~10 screens- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
- 09
- 10
04 · The decision
More people finished asking for a loan.
74%
of users who started the application completed it
~20%
fewer drop-offs across the flow
Measured before and after in the marketing team’s funnel reporting. 74% is a completion rate; ~20% is a reduction. They are different kinds of number.
05 · Act 1 result
People applied once, then never opened the app again.
Loans are infrequent by nature. There was nothing to come back to, and an app nobody opens can’t earn beyond a single transaction.
The proposal on the table: one rotating banner.
- Wants a bigger loan
- Wants reassurance
- Wants to invest
- Wants deals
06 · New problem
Advertising people choose to open is a product problem.
Instead of a banner, a categorised ecosystem, where the commercial inventory becomes useful surfaces rather than interruptions. It was my proposal in response to the banner request.
A single banner serves one interest to an audience with many.
Categorised surfaces mean every user has something relevant to open.
Relevance beats repetition: same inventory, placed where it’s wanted.
Ten surfaces. All designed solo, all shipped.
- Fundu Cash
- Fundu Card
- Fundu Club
- Fundu Shop
- Fundu Marketplace
- Fundu Trade
- Fundu Health
- Fundu Games
- Refer a Friend
- Upload Documents
Sell and buy near you
Open a demat account
Health checkups at home
Points become real cash
same inventory, now something you’d actually open
07 · Banner → ecosystem
100+ screens, one designer.
A tour of the shipped Android app, grouped by what people came to do. Swipe a row, tap any screen to open it full size.
Shipped 2020–21 · visuals refreshed 2026Up close: the home screen
Home & credit
3 screensLoan journey
6 screensEcosystem
6 screensRewards
3 screens08 · The UI, up close
Fundu Games. My idea. It failed.
The thesis: loans are infrequent, so manufacture a reason to return. Play games, earn points, redeem them. Then the ecosystem shipped, and people had genuinely useful reasons to come back.
Manufactured engagement loses to real utility.
What I did with it instead
Coins became the currency for offers: earned through play, spent on deals.
Paces inventory
Without it, a user clears every offer on day one and finds an empty screen the next.
Creates earned value
An offer you unlocked reads as worth more than one that was just sitting there.
“Wasn’t that just friction in front of ads?”
Playing was quick and optional, and the loan product was never gated. It only applied to the deals layer, which would otherwise empty in a single session.
09 · What I got wrong
From a one-time tool to an app with a reason to return.
74%
application completion after the 27 → ~10 restructure
~20%
fewer drop-offs
10
ecosystem surfaces shipped, solo
The app also reached 3M+ installs in its first year. That was driven by the parent company’s performance-marketing engine, so it’s company context, not a design claim. What happened after install was mine.
One profile, ten lenders
Users stopped repeating themselves.
A reason to return
A single-transaction product became a recurring one.
A second revenue line
Advertising inventory found a home users would actually open.
10 · Impact
What I’m taking with me.
Ambiguity is the design job
“Compete with a market leader” only became a problem because someone defined it.
Real utility wins
Real utility beats manufactured engagement. People return for things worth returning to, not for points.
Ask once, ask late
Collect the shared core one time; defer everything specific until the user is committed.
Evidence earns decisions
With no formal process and no authority, campaign data and funnel numbers moved the arguments.
11 · Takeaways
Thanks for reading.
Same instinct, different world: let the evidence decide. Next, a project where the research rewrote the brief.
Decide, don’t plan.
We were asked for a meal planner. The research said people don’t plan, they decide. So I designed for the decision.
57%of 99 surveyed don’t plan
51screens, solo