Matthew Voshell

Now that AI can build almost anything, the hard question is what should be built, and why. I have spent 12+ years answering it as an ethnographic product designer. I go to the people the product is for before anyone draws a screen, and what I find usually changes what gets built.

View work

View work

Book a 30-minute call

Book a 30-minute call

The work

What everyone assumed, who I went to, what that changed, and what happened.

Detectica

Lead UX Designer

2016

The compliance platform nobody would let me research

My whiteboard of interview notes from the compliance officers, headed Do not erase

Deployed in

82
countries
42
languages

What changed

Speed turned out not to be anywhere on the main list of user needs. The story map built without research was largely disproved. The product became person-centric, not alert based. It was built around how a compliance officer works through a case, not how the system raises alerts.

View study

Matchrox on three iPhones: a race plan with hand-offs, a race debrief and a race-day packing list

Matchrox

2025 to now

The operating system for HYROX doubles teams

An iPhone app I co-founded, designed and largely built. Find a doubles partner by run pace, station strengths and racing style, or bring the one you race with. Invite them with a race goal and see a projected joint finish. Plan all eight runs and eight stations together, share training, and message in a thread that translates but keeps HYROX terms. Run race week from two checklists, then review the result. In public beta, with Spaces for gyms in build.

View website

Matchrox on three iPhones: a race plan with hand-offs, a race debrief and a race-day packing list

Matchrox

2025 to now

The operating system for HYROX doubles teams

An iPhone app I co-founded, designed and largely built. Find a doubles partner by run pace, station strengths and racing style, or bring the one you race with. Invite them with a race goal and see a projected joint finish. Plan all eight runs and eight stations together, share training, and message in a thread that translates but keeps HYROX terms. Run race week from two checklists, then review the result. In public beta, with Spaces for gyms in build.

View website

The rental app on three iPhones: an income chart, a property overview with a lease renewal, and a move-out checklist

Rental Property Management

2022

Rental owners are not one customer

Most residential rentals are side income for people with other careers. I found 181 owners through Craigslist listings, contacted 179 and interviewed 15, across 13 cities, with the same 23 questions. They differ by where they are in ownership: getting started, growing, building, inheriting, or running a multi-generational business. Each acquires, operates and exits differently. My job became defining what the product should offer them.

View study

The rental app on three iPhones: an income chart, a property overview with a lease renewal, and a move-out checklist

Rental Property Management

2022

Rental owners are not one customer

Most residential rentals are side income for people with other careers. I found 181 owners through Craigslist listings, contacted 179 and interviewed 15, across 13 cities, with the same 23 questions. They differ by where they are in ownership: getting started, growing, building, inheriting, or running a multi-generational business. Each acquires, operates and exits differently. My job became defining what the product should offer them.

View study

M360 iPad app dashboard with sales goal gauges, volume bars and a map of the next account

Budweiser

2017

The best and the worst salesperson

The first M360 iPad app, built by another agency, was “not usable.” I went to a Bronx warehouse to interview two salespeople: the best and the worst. Both tracked accounts in spreadsheets, with no view of the monthly goals their pay depended on. I wrote the story map and put the data they needed in front. It launched and sales reps used it.

View study

M360 iPad app dashboard with sales goal gauges, volume bars and a map of the next account

Budweiser

2017

The best and the worst salesperson

The first M360 iPad app, built by another agency, was “not usable.” I went to a Bronx warehouse to interview two salespeople: the best and the worst. Both tracked accounts in spreadsheets, with no view of the monthly goals their pay depended on. I wrote the story map and put the data they needed in front. It launched and sales reps used it.

View study

Same pattern

eNCPT on three iPads: the home page, a grid of nutrition term cards and a step overview

eNCPT

2022 to 2023

Dietitians could not find the right term

Dietitians in hospitals, private practice and state government told me they could not find the right term while documenting. I redesigned search, navigation and term pages around the four steps of the Nutrition Care Process, the process they document. The Academy of Nutrition and Dietetics approved it.

View study

eNCPT on three iPads: the home page, a grid of nutrition term cards and a step overview

eNCPT

2022 to 2023

Dietitians could not find the right term

Dietitians in hospitals, private practice and state government told me they could not find the right term while documenting. I redesigned search, navigation and term pages around the four steps of the Nutrition Care Process, the process they document. The Academy of Nutrition and Dietetics approved it.

View study

SprintZero on four iPads: the story map, research participants, the team huddle and interview practice

SprintZero

2018 to 2023

Define what to build before building it

As founder I interviewed product managers, designers, researchers and engineers and mapped what they needed. We built a generative AI web app around that map. Funding ran out in 2023 and I shut it down. The pitch: three or four people working with AI could lap a large product team. Matchrox is me working that way.

View study

SprintZero on four iPads: the story map, research participants, the team huddle and interview practice

SprintZero

2018 to 2023

Define what to build before building it

As founder I interviewed product managers, designers, researchers and engineers and mapped what they needed. We built a generative AI web app around that map. Funding ran out in 2023 and I shut it down. The pitch: three or four people working with AI could lap a large product team. Matchrox is me working that way.

View study

A vendor product card with a five-step approval pipeline: Preliminary, CIO, CTO, PM, Final

Bank of America

2017 to 2018

One problem, eight organizations

For a few years, meetings about managing technology vendors ended with no permanent fix. It looked like one team’s problem. I ran ethnographic interviews with the employees who had it. They sat in about eight organizations inside the bank. So I worked with engineers and product owners to set up data pipelines from across the firm.

A vendor product card with a five-step approval pipeline: Preliminary, CIO, CTO, PM, Final

Bank of America

2017 to 2018

One problem, eight organizations

For a few years, meetings about managing technology vendors ended with no permanent fix. It looked like one team’s problem. I ran ethnographic interviews with the employees who had it. They sat in about eight organizations inside the bank. So I worked with engineers and product owners to set up data pipelines from across the firm.

Where it started

My internship concept on a monitor: an alert inbox of SEC filings beside a filing with confidence rings

S&P Global

2015

An alert inbox for SEC filings

As an intern I redesigned an SEC-filings alert platform as an inbox: each filing with a confidence score that analysts could agree or disagree with. I worked from conversations with engineers, not active users. When data from active users came back, the work went in another direction. I have not designed that way since.

My internship concept on a monitor: an alert inbox of SEC filings beside a filing with confidence rings

S&P Global

2015

An alert inbox for SEC filings

As an intern I redesigned an SEC-filings alert platform as an inbox: each filing with a confidence score that analysts could agree or disagree with. I worked from conversations with engineers, not active users. When data from active users came back, the work went in another direction. I have not designed that way since.

The proposed salon log-in screen on an iPad. The app was not built

George Mason University

2011 to 2012

A salon chain, a team of six, one year

My year-long senior capstone: I led a six-person team for a 15-location salon chain. I had pitched a custom iPad app. Customer polling and a teammate’s day in the shop showed waits were fine; the register and inventory were not. We configured an off-the-shelf system for the pilot salon, made training videos and presented it in May 2012.

The proposed salon log-in screen on an iPad. The app was not built

George Mason University

2011 to 2012

A salon chain, a team of six, one year

My year-long senior capstone: I led a six-person team for a 15-location salon chain. I had pitched a custom iPad app. Customer polling and a teammate’s day in the shop showed waits were fine; the register and inventory were not. We configured an off-the-shelf system for the pilot salon, made training videos and presented it in May 2012.

The gym app home screen on an iPhone: next class, cart items, nearest location and membership

UC San Diego

2015

Pumping Silicon

Pumping Silicon was the first app I designed, for my UX certificate at UC San Diego: next class, cart, nearest gym and membership on one home screen. I wrote the feature list from my own routine as a member.

Read article

The gym app home screen on an iPhone: next class, cart items, nearest location and membership

UC San Diego

2015

Pumping Silicon

Pumping Silicon was the first app I designed, for my UX certificate at UC San Diego: next class, cart, nearest gym and membership on one home screen. I wrote the feature list from my own routine as a member.

Read article

Additional work

Property Scout: two property cards, and the panel that lists what is in favor, what to watch and what is not known for one home

Property Scout

2026 to now

Property Scout

A triage app for residential investors: pick a market, scan a scored pipeline of single-family homes on a map, and star what’s worth a closer look.

Property Scout: two property cards, and the panel that lists what is in favor, what to watch and what is not known for one home

Property Scout

2026 to now

Property Scout

A triage app for residential investors: pick a market, scan a scored pipeline of single-family homes on a map, and star what’s worth a closer look.

Two Figma variable tables: brand colours behind, and theme colours that point to them in front

Disney

2024 to 2025

Organizing Figma variables

A design systems contract at Disney. I mapped how our Figma variables should be organized: collections for text, device, brand and theme, and how each one feeds the next.

Two Figma variable tables: brand colours behind, and theme colours that point to them in front

Disney

2024 to 2025

Organizing Figma variables

A design systems contract at Disney. I mapped how our Figma variables should be organized: collections for text, device, brand and theme, and how each one feeds the next.

The CARA prototype on two iPhones: the home screen with tasks, and the metrics screen with a rule coverage chart

EY

2019

Compliance and Regulatory Assistance

Mapping new regulations to a bank’s policies was manual, costly and error-prone. Our team took a winning innovation idea from story map to a clickable prototype for compliance officers.

The CARA prototype on two iPhones: the home screen with tasks, and the metrics screen with a rule coverage chart

EY

2019

Compliance and Regulatory Assistance

Mapping new regulations to a bank’s policies was manual, costly and error-prone. Our team took a winning innovation idea from story map to a clickable prototype for compliance officers.

The CARA prototype on two iPhones: the home screen with tasks, and the metrics screen with a rule coverage chart

EY

2019

Compliance and Regulatory Assistance

Mapping new regulations to a bank’s policies was manual, costly and error-prone. Our team took a winning innovation idea from story map to a clickable prototype for compliance officers.

See all work

How I work

Ethnographic user research is the observation and analysis of human groups, about five people per persona, to learn about their goals, needs, and frustrations.

My process as a decision flow of five questions. Is the team clear and aligned on the problem? If not, kickoff with stakeholders. Have we identified who is affected? If not, sync with sales and support, vet the knowledge base and run initial user interviews. Is the team aligned on the target personas? Then synthesized personas. Have all processes been identified? If not, analyze transcripts, create and validate user journeys. Are business and user goals defined? If not, workshop epics and define acceptance criteria. It ends in a user story map.

How I lead

Setting the team up for success.

One team

The best ideas win, and failure is a normal part of the process.

If they fail, I failed them

It is my responsibility to put people in a position to be successful.

Protect but push

Empower, and step in only when necessary.

“Good enough” is never good enough

Set the bar for good, and keep raising it.

Never ask for what I won’t do myself

Whatever I ask of the team, I am willing to do alongside them.

No band-aids

Identify the root problem, act, then hold a post-mortem so it does not recur.

Beyond the work

Photography, a podcast, and music I make in my free time.

Photography

I discovered creativity through a Nikon D600 with a 35mm f/1.4 lens.

Podcast

Conversations with designers about the careers they left and how they found their way into design.

Music

Demos I post in my free time.

Based in

Seattle, WA

Based in

Seattle, WA

Local time

Local time

Weather

Weather

Mount Rainier

Mount Rainier

Common questions

Do you still build?

Why are most case studies private?

How do you work with teams?

Let’s talk

I’m looking for contract work. Have a product to figure out, or want to read a private case study? Send a message, or book a call.

Book a 30-minute call

Asking for a case study? Say which one. I usually reply within 1 to 2 days.