Case study2021 → nowFounding member, DCX

Building the DCX team

We started as three new designers nobody listened to. Step by step we became the team that backs design with data, and I set up the framework behind each step.

Founding memberDCX (Digital Customer Experience), Bajaj Finserv
8Frameworks connecting design and data
1 month → 7 daysResearch turnaround, with rapid prototyping
Analytics · insight · testing · researchWhat I set up
How the team was built · one capability at a timeRuns on a framework I set up
  1. Three designers, unheardAbout a year of good work going nowhere
  2. The pitchA team that backs design with data
    How data reaches design · how libraries are made
  3. AnalyticsOpened to designers
    Google Analytics · ContentSquare · Medallia
  4. InsightsOur word started to count
    From data to insight
  5. A/B testingChanges tested on live traffic
    Adobe Target · CleverTap
  6. User researchResearchers hired, rules set
    Panels · agencies
  7. Rapid prototypingA month of research, in 7 days
    Site replica
ThroughoutEvery change tied to a business number
ThroughoutEverything written into the playbook
Each step added a capability on top of the last. Every one ran on a framework I set up.
01

The start

We were three new designers, and nobody listened.

We came straight from college, eager to make things. We did our best work for about a year, but barely anyone listened, and anything new was hard to get approved.

02

The backing

Then my boss pitched a team that backs design with data.

My boss saw the value in what we were saying and proposed a team that backs every design decision with data. I set up what that team would run on: the frameworks and processes for how data reaches design, and how our libraries get made. My boss took it to leadership.

DCX =

every digital interaction a customer has with our website and apps, and the impression they walk away with.

Measured by
Task success · error rate · effort score · usability
DCX team
12 → 23 people
Promise
3 big journeys a year

"Customer-centric view > technology. Customer behaviour > strategy."

Our first plan for DCX
03

Analytics

First, I gave designers access to the analytics.

Google Analytics was already in the company, but not in designers' hands. I opened it to them, with a process for using it. Then we added ContentSquare, which is built for design: clicks, misclicks, how people actually move through a page. Medallia and voice-of-customer surveys added what people said: 47,172 responses in January 2023 alone.

Dashboards like DCX Metrics let a designer check a page before redesigning it, and a product owner check it after.

The DCX Metrics dashboard in Looker Studio: a sidebar of reports (segmentation, page analysis, click-through rate, search analysis, journey analysis), user summary tiles, age and gender donuts, and city tables.
DCX Metrics. One report per question a designer asks: who, which page, what they clicked, what they searched, where the journey broke.
04

Insights

Then I wrote a way to turn data into insight.

Numbers alone don't change a design. I wrote a framework for reading them: audit the journey, analyse the data, benchmark, and only then redesign and test. Our proposals started getting approved, because each one came with its reason.

  1. 01Heuristic auditThe journey, scored against Nielsen's ten heuristics.
  2. 02Data analysisWhere people drop, click and misclick.
  3. 03BenchmarkingHow the best in class solve it.
  4. 04RedesignOne change per problem, tied to its data.
  5. 05TestLive A/B, or a rapid prototype.
  6. 06Best journeyThe winner becomes the standard.
05

A/B testing

Next, we started A/B testing on the live site.

I brought in Adobe Target, showed the business what it could do, and set up the system and its rules: what can be tested, on how much traffic, and how every result reports back to CleverTap and Google Analytics. Then we taught the product teams to propose their own tests: a problem, the data behind it, a hypothesis, one variant.

A · Control

Control: a 'Pay your bills and recharge' card with five shortcuts and a View All link.

What the data showed: people looping homepage → bill payments → login.

B · Variant

Variant: the same card with a 'My Bills' link and a View All tile.

The experiment: give them "My Bills" directly. "View all" still goes to bill payments.

A real experiment pair. One change, from one piece of data.
Traffic
5–25%of live visitors
Runs
~18 dayson the live site
Reports to
CleverTapand Google Analytics
Results in
48 hourswith revenue impact
If it wins
100%of traffic gets it
06

User research

Then we hired researchers to find out why.

A/B tests say which version wins, not why. Once the numbers had earned us a hearing, we added qualitative research: we hired researchers, brought in agencies, and set the rules they'd work by. I wrote our rules for remote usability testing, from goals and recruitment to the agency brief, and the UX audit checklist.

15people make a decent qualitative test
30make a good one, for interviews
100+before a survey counts
Colour-blind participantsrecruited wherever we can
07

Rapid prototyping

Research went from a month to 7 days.

Live tests can't try what isn't built. So we built a replica: a progressive web app (a website that installs and behaves like an app) that looks like the real site, wired to the same analytics, tested by a panel matched to real customer profiles. Surveys and moderated tests that used to take a month came back in a week.

Decision

Test in two places

Chose

Live A/B tests on the real site, and rapid prototypes in a replica that risks nothing.

Gave up

Realness. Panel users match real customers but aren't them; prototypes ran on mobile only, and only the tested path worked.

Because

A live test needs real traffic and weeks, and can't try new form steps or anything unbuilt. A prototype tells us why, not just which.

08

Business numbers

Every change was tied to a business number.

No change shipped on taste. Each one was tied to a number the business already cared about: task success up on this page, so the funnel improves by this much. That kept us moving, and earned us room to do more.

Design got a number of its own: DXS, a 0–10 digital experience score, with a target for every product.

09

The playbook

We wrote all of it into a playbook.

Every framework was written down so anyone on the team could run it. Tools got the same treatment: I mapped our tools phase by phase, scored 17 documentation tools and picked FigJam for almost everything, because it came with the Figma licences we already paid for. Notion sat beside it, and XMind handled information architecture (IA): how the pages are grouped and linked.

01Data, on-site

  • ContentSquare
  • Google Analytics
  • Usabilla
  • Medallia

02Research, off-site

  • Optimal Workshop
  • SurveyMonkey
  • UserTesting / UserZoom

03Business data

  • Power BI
  • Looker Studio

04Research repository

  • Dovetail
  • Notion

05Journeys

  • UXPressia
  • Smaply

06Design + handoff

  • Figma
  • Sketch
  • Zeplin
  • InVision Studio
  • Principle

07Documentation

  • FigJam
  • Notion
  • XMind, for IA
Partners
UX auditScreenroot · Techved · Upwork freelancers
Remote testingModulus · Peepal Design
Panel buildModulus
The tool and partner ecosystem, by phase. What we used at each step, and who we hired when we couldn't do it ourselves.
10

The seat

By FY24, design had its own column in the business plan.

In the FY24 plan for the consumer web business, "Own the experience" sits beside 450 million in traffic and ₹14,000 crore (₹140 billion) in business. Design was no longer asking for a seat. It had a number to hit, like everyone else in the room.

Design's column sits beside traffic and revenue.

Follow one change from data to playbook.

A real experiment on the bill-payments card, from the data that started it to the template it left behind.

Live · try it
01 · Data

Customers were going round in circles.

The journey data showed a loop: homepage, bill payments, login, and back to the homepage.

ContentSquareGoogle Analytics
What we saw
A loop: homepage → bill payments → login → homepage.
Why
Their own bills sit behind login. The card only offered a list of billers.
Hypothesis
A direct "My Bills" link, through login, ends the loop.
02 · Insight

They were looking for their own bills.

The insight framework asks three things: what we saw, why it happens, and the one change that should fix it.

Insight framework
03 · Design

We added a direct link to My Bills.

"My Bills" takes people to log in and see their bills. "View all" moves into the row and still opens bill payments.

FigmaReady to build
04 · Test

Real visitors picked the winner.

Adobe Target split a slice of real visitors between A and B. Every click reported back to the analytics we already trusted.

Adobe TargetCleverTapGoogle Analytics
05 · Prove

A win has to move a business number.

Each test is judged on a success metric and its revenue impact. Only a winner goes live for everyone.

Success metricRevenue impact
Experiment proposalPlaybook template
Problem
Customers loop: homepage → bill payments → login.
Backing data
Journey analytics, customer feedback
Hypothesis
A direct "My Bills" link ends the loop.
Variant
Auto-layoutNamingNo groupsMarked for export
06 · Playbook

The way we ran it became the template.

Every team now proposes the same way: the problem, the data behind it, a hypothesis, and a Figma file that passes the handoff checks above. The next change starts back at data.

Experiments Playbook

11

Also in this work

A rapid prototype in the wild

A 50-person study of the deposit pages, run without a moderator: 81.6% task success and click heatmaps on the participants' own phones.

Click heatmap on the fixed-deposit page of the prototype.Click heatmap on the systematic deposit plan page of the prototype.
Rapid prototyping
Review gates

Six checkpoints from content review to post-build QA, with a shared defect log that marks whether an issue is a repeat.

Practice · QA
Library budget and licences

The first design-library build and the team's Figma licences both went into the FY22 design budget. That story continues in Bajaj Design Library.

FY22 budget
Kudos award

Awarded for "Data Analysis for Design".

Sep 2023
Google AnalyticsContentSquareCleverTapMedalliaAdobe TargetLooker StudioFigmaFigJamNotionMaze