BBajaj Finance4 min read
The map that became a decision engine.
A redesign of Bajaj Finance's Store Locator — reframed from a place-finder nobody used into a surface that helps customers decide whether a trip is worth making, and act on it without leaving the app.
- Role
- Lead Designer — Strategy, Research & Service Design
- Scope
- Discovery to in-store decision
- Platform
- Bajaj Finance app — six business lines
- Year
- 2026
- Team
- Design lead with product, engineering, operations, legal and marketing partners
- My authority
- Owned problem framing, research synthesis, IA and the shipped design direction
Challenge
A store locator taking 510K product searches a month bounced two out of three visitors, and fewer than 2% of app users had ever opened it.
Strategy
I reframed the journey from discovery to intent to decision, rebuilt the six states where customers decide to continue or leave, and moved ownership from marketing to commerce.
Results
Task success 24.6% → 60% and bounce down 65%, with lead generation up across business lines after release.
overview
Overview
The brief was one line: "The Store Locator bounces. Redesign it." The numbers backed it up — 510,000 product searches a month routed to the locator, a 67% bounce rate, 24.6% task success, and every completed trip ending in Google Maps, where Bajaj could no longer see what happened.
But fewer than 2% of monthly active users had ever opened it. Before redesigning the interface, the question had to change: is the customer trying to find a location, or to accomplish something through one? The answer turned a marketing asset measured in page views into a commerce asset measured in conversion — with new ownership across five functions and six P&Ls.
We are the only party in the transaction who cannot see the transaction.
Three in four people who came here to find a store left without one.
research
Research
Twelve assumptions were written down before any research began. A clickstream audit of 993,000 users, 16 in-depth interviews with live tasks on production, 8 front-line agent interviews and 215 survey responses then tested them — six of them turned out to be inverted.
- Underused, not underperforming — 90% of customers never knew the locator existed. There were no complaints because there was no usage.
- Purpose-driven, not exploratory — 9 in 10 people arrive with one specific job and want it resolved in 30–60 seconds.
- The list decides, not the map — list views converted at 29–43%; the map converted 4–5× worse.
- Store and branch are one journey — customers use the words interchangeably, so the IA shouldn't separate them.
- What people wanted before going — phone number (67.5%), exact distance (65.9%), working hours (62.7%), in-stock confirmation (54.8%) and savings (42.1%).
Four archetypes came out of the synthesis — the Decisive Buyer, the Offer-Led Browser, the Monthly Verifier and the Branch Resolver — each with a different level of urgency and anxiety.
Twelve assumptions, written down before we looked
Writing the assumptions down first meant the research could genuinely disprove them, rather than confirm whatever the team already believed. Six came back inverted — including the one the brief was built on.
The most expensive assumption was that the locator underperformed. It did not: almost nobody had ever opened it. A feature 90% of customers do not know exists cannot be fixed by making it nicer, which is why the roadmap changed before a single screen did.
What people said they needed before travelling
The survey asked what would make someone confident enough to make the trip. The answers were unglamorous and specific: a phone number (67.5%), the exact distance (65.9%), working hours (62.7%), confirmation the product was in stock (54.8%) and what they would save (42.1%). Almost none of it was on the card people were expected to tap.
Four ways people arrive
Synthesis produced four archetypes, split across two very different emotional states. The Decisive Buyer and Offer-Led Browser arrive while shopping — receptive, open to an offer. The Monthly Verifier and Branch Resolver arrive while servicing something — often anxious, wanting one fact and no promotion. Designing one tone for both was part of what had gone wrong.
strategy
Strategy
The journey was reframed from Discovery (where is it?) to Intent (what's the job?) to Decision (is it worth going?) — and roadmap priority moved from polish to adoption, because a feature nobody knows about can't be made more usable.
- Intent first → ask the job, not the noun; a task-first IA replaced 11 categories that had a 46.4% mapping failure.
- Best path, not nearest place → route people to digital or physical fulfilment, whichever actually resolves the job.
- Never end at an address → every leaf ends in an action: apply, call, book a callback, pay.
- Honest absence → state what we don't know, with "last verified" timestamps instead of faking real-time data.
- Tone matches moment → no promotion on anxious, servicing-led surfaces.
Five cross-functional disagreements — storefront imagery, a central offers database, real-time offers, offer portability and map integration — were each resolved with a shipped compromise and an explicit record of what was given up.
Five disagreements, and what shipped
Strategy documents are easy; the decisions are where design leadership actually happens. Five cross-functional disagreements shaped what launched, and each one cost something worth recording:
- Real storefront photography. Shipped a phased image repository, verified photography on the highest-traffic stores first. Gave up a visually consistent list in phase one.
- A central offers database. Shipped distributed authoring with a shared store key and a mandatory validity window. Gave up speed, and permanent governance with it.
- Real-time offers. Shipped near-real-time with a "last verified" timestamp and a graceful fallback. Gave up the appearance of being live, which was the honest trade.
- Offer portability. Shipped a saveable reference plus call and callback. Gave up counter-side lookup, which moved to a later phase.
- Map integration. Shipped MapMyIndia with a list-first default and the map as a toggle. Gave up point-of-interest freshness, mitigated structurally rather than cosmetically.
Position before polish
The least popular recommendation was to stop optimising the interface and fix adoption first. It was also the one the evidence forced: with 2% of monthly actives ever opening the feature, a conversion improvement on a surface nobody reached would have produced a rounding error.
flows
Redesigned states
Instead of auditing 71 screens, the work focused on the six states where a customer decides to continue or leave — each rebuilt to carry product, category and pincode context through the whole journey.
Entry — show what's here before asking anything
An 11-category pin grid gave no sense of what was nearby. Replaced with named entity tabs, a stated count and radius, an editable location, and the product the customer came from kept in context.
Search — understand the way people actually ask
Search returned silence for anything it didn't match exactly. Added autocomplete from the second character, brand synonyms, pincode and vernacular queries, plus voice and photo input.
Results — list-first, with the answers people need to go
Cards had four equal-weight buttons and none of the information people asked for. Rebuilt list-first with a real storefront photo, rating, hours, distance and a branch-specific offer.
Decision — use what only Bajaj knows
The 20.5% card-tap rate wasn't a design defect — it was an ownership vacuum. The detail view now shows pre-approved limits, eligibility and offers from Bajaj systems, alongside phone, hours and ETA.
Trip — keep the journey inside the app
Every trip used to end with a hand-off to Google Maps. An in-app map with named pins, clustering and Apply / Call / Book callback actions now keeps the journey in Bajaj's app and captures visit outcomes.
Degraded — no dead ends
A denied location permission produced a blocking modal and a disabled button. Now there are three exit paths — allow, enter manually, or skip — and a wider radius when nothing is nearby.
The old journey: eleven category pins with no sense of what was nearby, cards with four equal-weight buttons and none of the information people asked for, and a hand-off to Google Maps that ended the relationship at the exact moment it mattered.
The redesigned journey: entity tabs with a stated count and an editable location, list-first results carrying distance, hours and a branch-specific offer, and a decision view that shows the pre-approved limit only Bajaj can know.
Three supporting states: search that accepts brand synonyms, pincodes and vernacular queries; an in-app map with named pins and next actions instead of a hand-off; and a degraded state with three ways forward rather than a blocking modal.
outcome
Outcome
What design drove, and what it did not. Task success and findability come from task-based validation on the redesigned states. Bounce, lead generation and business-line figures are production numbers measured after release, in a period that also included merchandising, offer and operational changes owned by the six business lines — the design work is one contributing factor, not the sole cause.
Time on task dropped from 87 to 60 seconds, and the effect carried into the businesses: store and dealer visits for B2B Retail EMI rose 25%, and Gold Loan business grew 10%. AI shortened research synthesis, audits and design variant exploration — but every shipped decision was checked against the evidence. The full walkthrough, including before/after screens and the service blueprint, is linked above.
What I would do differently
I audited the interface before verifying adoption. The field study that reframed the whole project happened in week four; it belonged in week one, and doing it in that order would have saved the team a month of careful work on the wrong problem.
I also drafted strategy before bringing all six business lines in. The offers contract turned out to be the critical path, and it surfaced later than it should have because the people who owned it were not in the room early enough.