Say hello Case overview
Case overview
NerdWallet · 2025–2026 · Full case article

Modernising how people match with a financial advisor.

A re-modernisation of the Advisors Match flow, end to end: from the first question to a booked consultation. Rebuilt mobile-first on a design system I built from scratch, and validated with A/B tests and real users at every step.

  • RolePrincipal Product Designer
  • TeamSole designer in the business unit
  • Worked withProduct, engineering, marketing, compliance
  • Year2025–2026
Live product · real screens
+27%
More leads generated
5
New sub-products in the experience
3+
Distinct journeys by type of user
87%
Of traffic on mobile
01Business context

Helping people find the right financial advisor.

NerdWallet Advisors Match connects people seeking financial guidance with advisors who may fit their needs, from a first expression of intent to a scheduled consultation.

NerdWallet already brought substantial traffic through SEO content and campaigns, and it sells leads. So the design focus was clear: convert that intent into qualified leads that reach the results page, while keeping trust in a high-consideration journey.

I was the Principal Product Designer and the only designer in the business unit, working with product, engineering, marketing and compliance.

Entry

SEO articles, the homepage or paid campaigns

Qualification

Profile, financial and goal questions

Personal info & verification

Contact details, phone code, consent

Matching

Advisors that may fit the user’s needs

Results & comparison

Firms side by side, with reasons

Advisor details

Approach, clients, minimums

Scheduling

Pick a time, confirm the free call

02The problem

The funnel lost most of its volume before anyone saw a match.

People come to Advisors Match to find an advisor, but the product has to qualify them first: ten steps of profile, financial and personal questions, then phone verification. All before a single match appears.

What the data shows

Volume halves at the ZIP code, drops again at the advisor questions, and keeps falling through every personal-information step.

Why it matters

Every lost user is a person who didn’t get advice, and a lead the business and its advisor partners never received.

Funnel progression by stepIndexed to the first question (Age = 100)
Age100
ZIP code52.9
Employment48.7
Marital status47.5
Annual income47.3
Advisor questions36
Name34.2
Email25.4
Mobile number19.7
Phone verification17

Largest single drop: the ZIP code, −47.1 points. Across the personal-information steps, 34.2 becomes 17.0.

Constraints & competing needs

Shorter wasn’t automatically better.

Cutting questions would raise completion and lower match quality. Four tensions pulled against each other, and each one shaped a design decision.

Conversion vs Lead quality01

Every question costs completion, but advisors need enough context for a useful first call.

Design implicationKeep the questions that drive the match; make each one cheaper to answer.
Trust vs Personal data02

Contact details are asked before any match is visible, in a category where people fear sales calls.

Design implicationGive people a concrete reason before the most sensitive step.
Speed vs Compliance03

TCPA consent and advisor disclosures are legally required before anyone can be contacted.

Design implicationDesign consent into the flow: readable, native, never a dead end.
One flow vs Many contexts04

SEO readers, homepage visitors and paid traffic arrive with different context, 87% of them on mobile.

Design implicationOne shared system, with journeys tuned per channel.

Three tensions, three design answers you can see on the screen.

The fourth, one flow for many contexts, is answered by the system further down.

Conversion vs Lead quality
Age question with its purpose and one-tap ranges
What we designed

Each question says why it’s asked, and answers in one tap.

Why it worksAdvisors keep the context they need; people spend less effort per answer.
Trust vs Personal data
Blurred matches shown as ready before the phone number
What we designed

Blurred matches appear “ready” before we ask for the phone number.

Why it worksA concrete reason to share the most sensitive detail, right when it’s asked.
Speed vs Compliance
TCPA consent as a short, explicit sheet
What we designed

Consent is a short, explicit sheet. The full legal text stays readable in place.

Why it worksCompliant without a dead end: people understand what they agree to and keep moving.
03Strategy

Four principles guided the redesign.

01

Show value before every ask

Make the outcome (matched advisors, a free call) visible before asking for effort or data.

02

Every question earns its place

Keep what drives the match; make each answer cheaper to give.

03

Trust and consent are part of the flow

Explain why we ask; disclosures and consent are readable, native and never a dead end.

04

Scale through one system

Every journey has to talk to the others and be designed as one. A shared system keeps them consistent as they multiply.

Principle 01 at the entry point

Expectations first, questions second.

The homepage sets expectations before the first question: three steps, free, no obligation. People know what they’ll get, and what it will cost them, before they spend any effort.

Same story on mobile, stacked, where 87% of the traffic lands.

NerdWallet Advisors Match · homepage
NerdWallet Advisors Match homepage, desktop
NerdWallet Advisors Match homepage, mobile
04How we learned

A/B tests and interviews, at the same time.

While the A/B tests ran, I validated the same variants in qualitative interviews. The numbers showed what people did; the conversations showed why.

Running both together made it possible to reach new discoveries, and new learnings, fast.

01 · Hypothesis

Grounded in the funnel data

Every change starts where people drop, and says what it should move.

02 · A/B tests

What people do

Variants run on live traffic, judged by leads reaching their matches.

03 · Interviews

Why they do it

The same variants, with real people talking them through.

Two directions, four open questions.

A mobile-first rebuild and a modernised funnel explored different answers to the same questions. Neither questionnaire was final; each one fed the tests that followed.

Intro
Mobile-first · Intro
Section break
Mobile-first · Section break
Age, scroll wheel
Mobile-first · Age, scroll wheel
ZIP, typed
Mobile-first · ZIP, typed
Stage stepper
Modernised funnel · Stage stepper
Stages + nudge
Modernised funnel · Stages + nudge
Grouped details
Modernised funnel · Grouped details
Context first · loading before contact details
One question per screen?Would focused, native inputs beat a long form on mobile?
How should progress feel?Stages that explain the journey, or one bar that feels close to done?
How much per screen?Fewer, denser screens, or more, lighter ones?
Context before a sensitive ask?Would a loading screen that explains the match make people readier to share contact details?
05Hypotheses & validation

Every decision started as a hypothesis.

Each one was grounded in the funnel data, and only became part of the flow after real users confirmed it. Three of them changed the shape of the questionnaire.

HYPOTHESIS 01 / 03

Remove effort from the ZIP code.

HypothesisIf we offer “Use my current location” and confirm the city inline, fewer people will drop at this step.
EvidenceZIP code was the largest single drop in the funnel (100 → 52.9).
ValidatedTested with real users; location assist and inline validation became the standard ZIP step.
Typed, keypad only
BeforeTyped, keypad only
AfterLocation assist + city check
HYPOTHESIS 02 / 03

Show value before the phone ask.

HypothesisIf people see their matches are ready before giving a phone number, more will complete the most sensitive step.
EvidenceVolume halves across the personal-information steps (34.2 → 17.0).
ValidatedTested with real users; the blurred matches preview stayed in front of the phone step.
A bare request
BeforeA bare request
AfterMatches ready, blurred
HYPOTHESIS 03 / 03

Lead with the user’s goal.

HypothesisIf the flow opens with what the user wants, it will feel like advice, not a profile check.
EvidenceThe flow opened with demographics; goals came later, if at all.
ValidatedTested with real users; the goal question now opens the questionnaire.
Opens with age
BeforeOpens with age
AfterOpens with a priority
06Testing & learning

HUE, in practice.

Traffic was already strong, so conversion was the question. To guide decisions while iterating, we built HUE (Happiness, Usability, Engagement), our own adaptation of Google’s HEART framework.

Here are three of the experiments: what we tested, what we observed, and what we chose.

H
Happiness

Progress that feels close to done.

ExperimentOne single non-linear progress bar vs. a three-stage indicator.
ObservedIn qualitative sessions and funnel tests, the single non-linear bar performed better than the three-stage indicator.
Our readingIt felt closer to done; stages made the journey feel longer.
Three-stage indicator
Explored

Three stages

Chosen

One bar

U
Usability

An age picker that takes one tap.

ExperimentScroll wheel vs. text input vs. selectable ranges that advance on tap.
ObservedSelectable ranges that advance on tap had stronger user acceptance and better completion than scrolling or typing.
Age, scroll wheel
Tested

Scroll

Age, text input
Tested

Text input

Chosen

Ranges, one tap

E
Engagement

Legal texts that build trust, not friction.

ExperimentThe same legal and disclosure text, shown open in the middle of the screen vs. behind a shield icon at the top that opens a native sheet.
ObservedPeople who opened the legal texts from the shield icon trusted the flow and converted more than people who met the same text open in the middle of the screen.
Our readingA wall of legal copy mid-form reads as risk. Kept a tap away, behind a security icon, the same text reads as transparency: the screen stays focused, and those who want the detail get it on their own terms.
Legal text open mid-screen
Tested

Open mid-screen

Chosen

Behind the shield

Fewer screens didn’t mean less effort.

Grouping name, email and phone on one screen looked like the obvious shortcut. We tested it against one detail per step.

One grouped screen
Converted worse

A · One grouped screen

Name, last name and email together.
HypothesisGrouping name, email and phone on one screen would reduce the apparent number of steps and simplify the form.
LearningSeparating the details converted more effectively than collecting them together.
DecisionOne detail per step: a lighter ask each time, in a clear progression into verification and consent.

1 · Name

2 · Email

3 · Mobile number

4 · Verification

Converted better

B · One detail per step

Each with a short reason, and the matches preview before the phone ask.
07The solution

Every decision, in motion.

The shipped flow end to end, built from the real screens. The journey runs in four moments, and every change in it was a hypothesis, tested with real users and measured.

01 · Goals first

Lead with the goal, answer in a tap.

Goals, finances and life stage. Each question says why it shapes the match, and answers in a tap.

Shipped · Live product

Top priority

The goal opens the flow

ZIP code

Location assist + city check
02 · Contact details

Earn the most sensitive ask.

The most sensitive ask needed the most convincing: a loading screen that explains the match, blurred matches as a trigger, and readable legal text.

Loading screen

Explains what happens next

Matches as a trigger

Value before the phone ask

Readable consent

TCPA, explicit and in place
03 · Matching

Matches people can compare, and act on.

After verification, the job changes from collecting information to helping someone choose. Each match ties the user’s answers back to the result; firms sit side by side with the same structure; specialities, assets under management and a verified badge show trust before any commitment.

Your matches

Each firm with why it was recommended

Compare

The same structure, one swipe away
04 · Booking

From a chosen match to a booked call.

The lead is generated when the user reaches their matches. The last moment gives them a clear next step: a free consultation in a few taps. Booking lives inside the match, so the user never leaves the context that convinced them. It later grew into a sub-product of its own.

Advisor detail

Approach, clients, minimums

Pick a time

The advisor’s calendar, inline

Booked

Who, when, add to calendar

Live product. Screens show the shipped flow at the time of the work; ongoing A/B tests and variations may differ from what is live today.

08Built from scratch, AI-first

One system, many journeys.

Three entry points, tailored journeys and five sub-products, with 87% of traffic on mobile. Every journey had to talk to the others, so they had to be designed as one.

I built the design system from scratch, on my own: foundations, components with their states, and compositions reused across mobile and desktop. It was AI-first from documentation to implementation, and I had full autonomy over how components were used and developed.

  1. FoundationsColour by role, type by job, one layout grid.
  2. Components with their statesActions, inputs, selection and guidance: enabled, disabled, selected, valid.
  3. Documentation, AI-firstWritten and kept up to date with AI.
  4. Implementation, AI-firstFrom the documented component into code.
  5. Full autonomyOver component use and development.
Foundations: colour by role, type by job, layout grid Actions and inputs with their states Selection and guidance patterns Orientation: the same frame on every step of every journey

New journeys start from tested components, and a familiar frame on every step reassures people. That is how one flow could serve SEO readers, homepage visitors and paid traffic without becoming three different products.

09Results

More qualified leads, without trading away trust.

+27%

More leads generated: people who reach their matches after verification. Of every 100 who start, 21.6 now get there, up from 17.0.

+5.8 pts

More starts past the ZIP code, the biggest leak.

Stable

Advisor-rated lead quality vs. baseline.

First question100 → 100
ZIP code52.9 → 58.7
Employment48.7 → 54.1
Marital status47.5 → 52.7
Annual income47.3 → 52.5
Advisor questions36 → 41.2
Name34.2 → 39.7
Email25.4 → 30.5
Mobile number19.7 → 25
Lead · reaches matches17 → 21.6
BeforeAfterRedesigned step
Business
−21%

Cost per lead on paid campaigns: same spend, +27% leads.

Happiness
4.2 → 4.6

Satisfaction on the results page (CSAT, out of 5).

Usability
−24%

Median time to finish the questionnaire.

Engagement
+14%

Completion among users who opened the disclosures sheet, vs. inline text.

10Reflection

Trust and conversion aren’t a trade-off.

In regulated, lead-based products, good design is an honest exchange: ask only for what you need, explain why, and give something back at every step. Done well, trust compounds into conversion.

The size of the ask beats the number of screens.

Separate, explained steps converted better than one grouped form. People don’t count screens; they weigh what each one asks of them.

Conversion and lead quality are one problem.

Every question removed had to be weighed against the match it supports. Growth that sends advisors worse leads isn’t growth.

Measure the journey, not the step.

A lift at one step can hide a loss further on. We judged every change by leads reaching their matches, not by the step it touched.

Thanks for reading. Next up: Doctoranytime.

From a doctor directory to a broader health ecosystem: a new brand, a complete product experience and a design system with tokens built for the business.