Case study2021–25Design lead, team of 3HEROES award · Apr 2022

BFL Website Redesign

The site gets 24.3M visits a month. I looked for where people gave up, and fixed what made them leave, one problem at a time.

+21.6%conversion
+23%traffic
24.3Mvisits a month
6 · 24 · 3products · journeys · people on the team
After · the product page
01

Where they left

A site this big loses people in small places.

Six products, 24 journeys, three of us. Redesigning it all wasn't possible, so instead of asking which pages looked old, I asked where people left. The analytics showed people dropping off in many places across the site. We found the specific problem behind each one and worked through them one by one.

FD form · mobile verification

86.2%left at the very first step
  • 86.2%exited the form
  • 8.1%went on to get the code
  • 5.7%went elsewhere

FD form · average time on form

Finished the form17:00
Gave up3:46

Finishing took 17 minutes. People who quit left before 4.

FD · conversion by device

Mobile0.14%
Other devices1.40%

Most traffic came from phones, where a visit was a tenth as likely to end in a deposit.

Customer portal · the five-step login

Started login13,373,793
Finished login8,371,89062.6% kept

More than a third of the people who started logging in never got in.

02

The constraint

Leadership wanted payment first. I designed around it.

This case follows one of those problems: the FD (fixed deposit) journey, which came with a constraint from leadership. The old journey asked for everything before taking any money. Payment came at stage 10, and 82% of people who reached it dropped off there. To win small savers, leadership set one rule for deposits up to ₹25,000: take the money first.

Before · 10 stagespayment at 10

  1. 1Investment details
  2. 2Personal details
  3. 3Primary holder KYC
  4. 4Email ID
  5. 5Income & declaration
  6. 6Joint holder & KYC
  7. 7Nominee details
  8. 8Bank verification
  9. 9Summary & T&C
  10. 10Payment
  11. ✓Thank you

After · 7 stagespayment at 3 · four stages after it

  1. 1Investment details
  2. 2Personal details, prefilled
  3. 3Payment
  4. 4Income & declaration
  5. 5Nominee details
  6. 6Joint holder & declaration
  7. 7KYC appointment
  8. ✓Thank you
A new customer, depositing up to ₹25,000.

To make that work, I sorted every field by why we were asking for it.

  1. Before paymentThey chooseAmount, tenure, payout. Now on the product page.
  2. After paymentThey may shareNominee, joint holder. Optional.
  3. After paymentWe must ask by lawKYC (the identity check) and declarations.
03

Choosing

Choosing an FD went from 49 seconds to 13.

With payment at stage 3, choosing had to be quick. Before building anything, we tested four product pages (PDPs) with 110 people. The winner put three ready-made options up front, with a custom field beside them.

MeasurePDP 1PDP 2PDP 3PDP 4
Task success71.7%79.1%86.7%88.9%
Average time49.1s46.5s27.8s13.2s
Found the highest rate easily57%86%68%92%
04

The new problem

Once people paid, they thought they were done.

Paying feels like the end. With payment at stage 3, four stages still stood between the customer and a booked deposit, and our prototype test showed how few people knew it.

30%

knew the steps after payment had to be done within 10 days

32%

knew their interest had started on the day they paid

05

The tests

I A/B tested the tone of the screen after paying.

Payment couldn't move, so I worked on the screen right after it. We ran A/B tests (versions shown to different visitors, then compared) on how much text, which colours, and what the message said.

Tone decided it. Red warnings worried people, and worried people stop. A big success screen told them they were finished. The version we kept read as routine: money in, interest running, two short steps left.

Trade-off

Calm over loud

Chose

A plain status on every screen after payment: money received, interest started, what's left.

Gave up

The loud warning. "Finish in 10 days or get a refund" stays, in one quiet line.

Because

In the tests, worried people stopped. Settled people went on to finish.

06

Errors

The quietest exit was an error nobody saw.

On the loan forms, people pressed submit 4.2 times on average. The error messages were there, scrolled out of view. Now submit jumps to the first error and shakes the field.

4.2→1.4

attempts per user to submit a loan form

32%→12%

error rate

07

Hindi and Tamil

Hindi takes 1.2× the room. Tamil takes 1.5×.

Many customers read best in their own language, so the site went into Hindi and Tamil. The same sentence runs longer, and every component built for English had to hold it. We tested with 10 Hindi and 10 Tamil readers, then reworked headings, tiles and buttons to fit.

Flexi Personal LoanGet a loan of up to Rs. 25 lakhs
  • Approval in 5 minutes
  • 24-hour disbursal
  • No collateral
EMI for 39 months at 13% p.a.

Need help in choosing the right Personal Loan?

Know MoreApply Now
English1×
Hindi1.2×
Tamil1.5×
  • Actions are translated.Know More becomes अधिक जानें.
  • Familiar words keep their English sound.Apply, Login and Payment are spelled out in the script: अभी अप्लाई करें.
  • EMI, KYC and CIBIL stay in English.Customers already know them that way.
  • Fonts matched to Rubik, the site's font.Baloo 2 for Hindi, Mukta Malar for Tamil.
08

The result

Four years of small fixes added up.

+21.6%

conversion across the redesigned journeys, with traffic up 23%.

  1. Journeys mapped

    Across products; the credit-card form audited screen by screen.

  2. FD form analysed · HEROES award

    17 minutes to finish, 86.2% exit at mobile verification.

  3. Customer portal mapped

    106.2M homepage visits; 62.6% finished the five-step login.

  4. Forms

    Attempts per user 4.2 → 1.4, errors 32% → 12%.

  5. Hindi and Tamil

    The site in two more languages.

  6. FD payment first

    Product pages and the screen after paying, tested.

  7. Coded prototypes

    The site in code: three builds, about 190 pages each.

“Exemplary work on digital journey design for various businesses.”

HEROES award citation, Bajaj Finserv, April 2022
09

Also in this work

Mutual Funds 2.0

I led a cross-functional team, took apart the pages of Groww, a leading rival, and set rules like “transaction first, then onboard”.

2024 · bajajfinserv.in
Gold Loan

One in four phone visits to the branch finder waited 10.6s or more. My A/B test loaded only the branches in your state.

A/B test · branch pages
Hamara Mall

In a test of Bajaj's local-shopping app, 9 of 10 people chose the busier version. To them, simple meant familiar, not empty.

User test · local-shopping app
Coded prototypes

Three builds of about 190 pages each, 21 calculators and every form state, so reviews happened in a browser. How we worked is in Building the DCX team.

Feb–Jun 2025 · HTML, CSS, jQuery
Google AnalyticsCleverTapContentSquareMedalliaAdobe TargetSketchFigmaHTML / CSS / jQuery