Three very different users. One rigid flow.
ShopSe is India's largest digital EMI marketplace: a point-of-sale Buy Now Pay Later platform powered by 10+ NBFCs and banks. The challenge was hidden beneath that scale: three fundamentally different user segments pushed through the same fragmented offline flow. High drop-off, merchant-heavy, and a hard ceiling on growth.
Pre-Approved Users
The bank already knows them. Credit limit pre-assigned. Fastest path: phone → offers → EMI → done.
Non-Pre-Approved (NPA)
No prior relationship with a lender. Real-time underwriting: identity, income, credit score, full KYC.
Credit Card EMI
Already holds a credit card. Pick a tenure, confirm the EMI on the existing card.
A checkout designed for the system, not the user.
Built incrementally around lender API sequences. Each partnership added screens. The flow worked fast for two segments but broke for the third. A 2-minute purchase became a 40-minute ordeal.
From the original flow · 7 of 25 screens
I went to the stores. Three things changed my approach.
A heuristic audit of all 25 screens, then contextual research in real stores: 5 live transactions shadowed at the counter and field notes from 12+ shops. Insights came from observation, not questions.
"This feels like a bank loan application, not a phone purchase."
NPA user · Session 3"Which one is actually cheaper overall?"
Asked to merchant · 3 of 5 sessions"Is this normal? Why so much scrolling?"
At Form 6 · 1140px scroll height"The goal isn't to remove friction. It's to move clarity earlier in the journey."
Design principle established after field researchTwo layers of decisioning. We were using the wrong one first.
In non-pre-approved flows there are two underwriting layers. The old flow forced the expensive full check before giving the user any signal at all. Understanding this infrastructure, not just the UI, is what unlocked the redesign.
Soft Check
Pre-eligibility- Inputs
- Phone number · PAN · Date of birth
- Output
- Directional signal: "You may be eligible for Bank A up to ₹50K." No hard credit impact.
- Speed / Accuracy
- Fast · Directional
Full Check
Final approval- Inputs
- Employment · Income · Bank account · Full KYC
- Output
- Binding approval with exact EMI plans and disbursement terms. Accurate but slower.
- Speed / Accuracy
- Slow · High
Four approaches. One winner.
Instead of jumping into UI, I explored four different ways to structure the journey and argued each one against the field evidence.
User selects a lender, fills the form, then sees eligibility.
Users don't know which bank will approve. Choice without clarity creates dead-ends.Collect every detail upfront, then show all eligible banks.
High upfront effort feels overwhelming. Effort before clarity.Show all banks, highlight "most likely approval."
Reduces confusion but doesn't eliminate repeated effort.User fills details once. System returns all valid lender options instantly.
Users accept effort when it guarantees a clear, comparable result.Eligibility-first, trust-aware checkout.
Four phases, each designed to answer the user's primary question at that stage.
Onboarding
"Who am I?"Single login screen with inline OTP. Eligibility runs silently.
2 screensSelection
"What can I afford?"Direct search or barcode scan. Side-by-side EMI comparison.
4 screensVerification
"Can I be trusted?"3 focused forms, email verification, Review Details checkpoint.
4 screensCompletion
"Is it done?"Mandate summary. GPay. Narrative loading states.
3 screens
Selected screens from the new flow · 13 unique screens total
EMI comparison: a new capability.
"Users couldn't tell from our list which plan was cheaper. The merchant did the math for them." This screen turned a settings-menu list into a financial decision surface. It is the hero moment of the redesign.
Lender partners pushed back on exposing comparable totals. I brought session recordings of users asking merchants "which is cheaper?" and the founder backed the user-first call.
Nine became seven. Errors dropped.
Five long forms became three focused forms grouped by purpose with plain-language sections and inline tooltips, plus email verification and a Review Details checkpoint before KYC submission. Three fields were removed after auditing what compliance actually required.
The checkpoint directly reduced rejections from mistyped PAN and Aadhaar details. Those errors used to end the journey entirely and burn a real credit inquiry.
Narrative loading states as trust design.
The old payment flow had a blank redirect: 5 to 8 seconds of white screen. Users described it as "sketchy." I designed three narrative loading screens with named progress stages to fill that silence with confidence. Waiting reads as motion, not failure.
Every word in the key message earns its place. Not "Loading" (vague). Not "Processing" (cold). "Matching your best EMI plan" is specific, personal, and working on your behalf.
V1: designer thinking. V2: systems thinking.
V1 taught me where friction lived. V2 addressed why it existed. Between the first pass and the shipped build, each area was rewritten by what we watched real users do in shops.
The conversations behind the calls.
Working solo meant every call was argued in review and settled with evidence. Five conversations that shaped the product, and how each one ended. None were won by volume. Each was won by showing, not arguing.
"Every commerce app browses by category. Users expect it."
Field notes from 12+ shops: users pointed at products by name. Not one navigated by category.
✓ Scoped search on the existing SKU system. Four steps became one.
"Why not AI-powered recommendations instead of plain search?"
No behavioral data to train on yet, and a wrong suggestion in a finance flow burns trust we did not have.
✓ Parked for a later version. Search shipped first.
"The EMI list just needs better styling."
Sessions showed users doing mental math across screens. Choosing a tenure is a financial decision, not a menu.
✓ Side-by-side comparison, with total cost and savings up front.
"The KYC form is what compliance requires. It cannot shrink."
Audited the form field by field against what was actually required.
✓ Three fields removed. The rest grouped into guided sections with tooltips.
"Customers say the app looks sketchy. They ask us for the paper form."
Treated trust as a design surface: one brand system, named progress stages, no silent waits.
✓ The trust layer shipped across every screen of the flow.
Process metrics are just the surface.
User behavior shifted
NPA users went from 20–40 minute checkout ordeals to a clear eligibility signal in under 60 seconds. Drop-offs moved earlier in the funnel: low-cost exits at login and search instead of expensive abandonment at Form 6 and KYC.
Operational cost dropped
The Review Details checkpoint caught PAN and Aadhaar errors before they hit lender APIs. Errors that previously triggered irreversible rejections became editable in-flow. Merchant training got simpler: fewer screens, fewer support tickets.
Online expansion unlocked
The architecture made the shift from offline-only to online viable. ShopSe expanded beyond Electronics into Health & Wellness (2× repeat customers), EdTech (+55% enrollments), and Insurance (+48% policy sales).
Shipped on the system · 85+ component design system across 4 product teams · WCAG 2.1 AA coverage on 47 flows
The shipped flow, login to receipt.
Eight moments from the production flow. Every screen below shipped to 200K+ monthly active users.
What this project taught me.
Want the full walkthrough?
I'm happy to walk through the research artifacts, iteration history, and decision trade-offs in detail, including the parts that didn't work.