The Michigan
Department of State

BACK

Overview

Problem

Research

Key Decisions

Final Design

Reflection

Michigan Department of State

Civic Technology & Data Visualization

Michigan Department of State

Civic Technology & Data Visualization

Michigan Department of State

Civic Technology & Data Visualization

Designing Clarity Into Michigan’s Election Results

Designing Clarity Into Michigan’s Election Results

Designing Clarity Into Michigan’s Election Results

Designing Clarity Into Michigan’s Election Results

I served as the Project Manager and Lead UX Researcher for the Michigan Department of State in redesigning their historical election results website, transforming a static archive of election tables into an interactive dashboard that makes official election data easier to explore and understand. This work expanded on the Center for Civic Design's Power BI Voting Dashboard. My team and I introduced the project at a statewide press conference alongside Secretary of State Jocelyn Benson and the Deputy Chief of Staff Neil Tambe.

🚀 Project is set to launch ahead of the August 2026 Primary Elections.

RECOGNITION

MDOS Press Release

Center Civic Design

UMSI Article

TIMELINE

Sep 2024 - Apr 2025

Sep 2024 - Apr

2025 (8 Months)

TEAM

1 Project Manager (Me)

2 UX Researchers

2 UX Designers

MICHIGAN RESIDENTS

7.3+ Million

7.3+ Million

active registered voters the dashboard is designed to serve

PERCEIVED DIFFICULTY

43%

43%

reduction in user-rated task difficulty across all tasks

TASK COMPLETION TIME

52%

52%

average reduction in time to complete key tasks

JURISDICTIONS

83 Counties

83 Counties

of Michigan election data unified into a single search interface

PROBLEM SPACE

Buried In Plain Sight

Government websites are rich in data but notoriously difficult to navigate. In an era of election misinformation, official data sources need to do more than provide data, they need to convey election information clearly and in context.

Michigan's existing historical election results website returns each election as one long, static page of tables. Finding a specific race means scrolling through hundreds of lines. There's no search bar. No visualizations. No context for what the numbers mean.

PROJECT GOALS

What We Set Out to Build

The Michigan Department of State came to us with a clear goal: make historical election results easier to find, understand, and trust. Building upon the existing Power BI Voting Dashboard, the redesign also had to work within the constraints of the platform. From that brief, my team established three guiding principles to anchor every research and design decision throughout the project.

PROJECT GOALS

What We Set Out to Build

The Michigan Department of State came to us with a clear goal, make historical election results easier to find, understand, and trust. Building upon the existing dashboard, it also has to follow the constraints laid out in Power BI. From that brief, my team established three guiding principles to anchor every research and design decision throughout the project.

PROJECT GOALS

What We Set Out to Build

The Michigan Department of State came to us with a clear goal: make historical election results easier to find, understand, and trust. Building upon the existing Power BI Voting Dashboard, the redesign also had to work within the constraints of the platform. From that brief, my team established three guiding principles to anchor every research and design decision throughout the project.

1

Improve Usability

Make the dashboard easier to navigate and understand

2

Increase Public Trust

Present election data clearly and in context to reduce confusion

3

Design For Scale

Build for scalability in Power BI to support past and future elections

RESEARCH

Navigating Political Stakes

Elections in a swing state carry real weight. My team understood that this dashboard could significantly shape how voters, journalists, and officials interpret Michigan's election data. Because of that, neutrality was the foundation of the project. To uphold that standard, we built accountability checks into our process.

  1. Client Alignment: Weekly syncs with Deputy Chief of Staff Neil Tambe ensured our design decisions aligned with what the Michigan Department of State could publicly support and defend.

  2. Faculty Critique: Regular stand-ups with academic advisors and peers surfaced blind spots and challenged assumptions before they became design choices.

  3. Neutrality by Design: Every design choice was evaluated through a nonpartisan lens with no biased framing, no editorial language, and no visual treatment that could favor a candidate or party.

Jocelyn Benson,

Secretary of State

“

Making those official sources of information more user-friendly, more accessible, and easier to understand makes them more useful. It prebunks misinformation and builds trust in our elections.

”

Michigan Voting Dashboard Press Conference, Oct 2024

RESEARCH

Information Architecture

Before proposing any solutions, we mapped the information architecture of the existing website to understand how election data was organized and presented. This analysis revealed a fragmented structure where users had to navigate long pages of static tables with little support for finding or interpreting information. Understanding these structural limitations gave us a baseline for evaluating future design decisions.

RESEARCH

Information Architecture

Before proposing any solutions, we mapped the information architecture of the existing website to understand how election data was organized and presented. This analysis revealed a fragmented structure where users had to navigate long pages of static tables with little support for finding or interpreting information. Understanding these structural limitations gave us a baseline for evaluating future design decisions.

RESEARCH

Information Architecture

Before proposing any solutions, we mapped the information architecture of the existing website to understand how election data was organized and presented. This analysis revealed a fragmented structure where users had to navigate long pages of static tables with little support for finding or interpreting information. Understanding these structural limitations gave us a baseline for evaluating future design decisions.

COMPARATIVE ANALYSIS

Existing Dashboards

We conducted a comparative analysis of election dashboards from other states and organizations to identify established patterns, usability strengths, and common pitfalls. These examples helped benchmark expectations for search, filtering, data visualization, and contextual guidance, helping us identify opportunities to improve Michigan's experience.

COMPARATIVE ANALYSIS

Existing Dashboards

We conducted a comparative analysis of election dashboards from other states and organizations to identify established patterns, usability strengths, and common pitfalls. These examples helped benchmark expectations for search, filtering, data visualization, and contextual guidance, helping us identify opportunities to improve Michigan's experience.

COMPARATIVE ANALYSIS

Existing Dashboards

We conducted a comparative analysis of election dashboards from other states and organizations to identify established patterns, usability strengths, and common pitfalls. These examples helped benchmark expectations for search, filtering, data visualization, and contextual guidance, helping us identify opportunities to improve Michigan's experience.

✅ Election Data Visualized

✅ Interactive Maps

🚫 Unintuitive Navigation

🚫 Limited Data Filtering

🚫 Lack of Context for Data

✅ Map & Pie Chart Visualizations

✅ Prominent Data Export Button

✅ Clear Update Timestamps

🚫 Data Heavily Text Based

🚫 Only Shows Past Election Results

✅ Clear Filtering & Navigation

✅ Data Source Clearly Labeled

🚫 Visualizations Are Static

🚫 Limited Data Views

✅ Clear Demographic Categories

✅ Data Is Contextualized

🚫 Color Contrast Issues

🚫 Static Charts & Filtering

✅ Highly Interactive & Intuitive

✅ Clear Visual Hierarchy

✅ Frequent Updates & Context

🚫 Less Local Level Reporting

USER RESEARCH

Understanding Our Users

I led user research with four primary stakeholder groups: voters, reporters, election clerks, and elected officials. Through interviews and journey mapping, we explored how each group interacted with election information. Although their goals differed, every group shared one need: quickly finding trustworthy election results and understanding what the data meant.

USER RESEARCH

Understanding Our Users

I led user research with four primary stakeholder groups: voters, reporters, election clerks, and elected officials. Through interviews and journey mapping, we explored how each group interacted with election information. Although their goals differed, every group shared one need: quickly finding trustworthy election results and understanding what the data meant.

🗳️

Voters

Want to find and understand what the data means to make decisions.

📰

Reporters

Want reliable, downloadable data to analyze trends and report accurately.

🧑‍💼

Clerks

Want to find and understand what the data means to make decisions.

👔

Elected Officials

Want flexible ways to view the data and visuals to share with the public.

View User Journey Maps

USER RESEARCH

Usability Testing

I moderated usability tests with 15 participants, asking them to complete seven representative tasks using the existing election results website. These tasks reflected common workflows. Across participants, we observed recurring breakdowns in navigation, terminology, and information hierarchy that made even routine tasks unnecessarily difficult.

USER RESEARCH

Usability Testing

I moderated usability tests with 15 participants, asking them to complete seven representative tasks using the existing election results website. These tasks reflected common workflows. Across participants, we observed recurring breakdowns in navigation, terminology, and information hierarchy that made even routine tasks unnecessarily difficult.

"People don’t understand that unofficial data is unofficial for a reason. Clear messaging about when counts are final would go a long way in building trust"

"It’s important to provide context with data visualizations…too often, maps and charts can mislead people if they don’t explain what’s behind the numbers."

"The current system feels built for the public, not for someone who works with election data regularly. There’s room to grow."

"Navigating the website is difficult [it] leads to Voter Information, then Michigan Voter Info Center, and then a different election results page. The way the sites are nested makes it hard to navigate."

"The current system feels built for the public, not for someone who works with election data regularly. There’s room to grow."

"I want a single election portal for all the interesting, valuable data that MDOS has to offer. I don’t want to scrape through 83 counties one by one."

"Download options in TXT or Excel for each county are essential. Having access to raw data is really important, and the ability to aggregate it for analysis would make the process much easier."

"I want a single election portal for all the interesting, valuable data that MDOS has to offer. I don’t want to scrape through 83 counties one by one."

"Download options in TXT or Excel for each county are essential. Having access to raw data is really important, and the ability to aggregate it for analysis would make the process much easier."

"Navigating the website is difficult [it] leads to Voter Information, then Michigan Voter Info Center, and then a different election results page. The way the sites are nested makes it hard to navigate."

View Usability Findings

USER RESEARCH

From Insights to Priorities

Following our research, we synthesized findings through affinity mapping to identify recurring themes across participants. We then used a priority-versus-feasibility matrix to evaluate potential features, balancing user impact with implementation constraints in Power BI.

USER RESEARCH

From Insights to Priorities

Following our research, we synthesized findings through affinity mapping to identify recurring themes across participants. We then used a priority-versus-feasibility matrix to evaluate potential features, balancing user impact with implementation constraints in Power BI.

Click to Zoom

DESIGN

Ideating With AI

We began ideation by using AI to rapidly generate a broad range of dashboard concepts based on our research findings and design requirements. These concepts were never treated as final solutions. Instead, they served as conversation starters that helped the team explore layouts and align on direction before proceeding to wireframing.

DESIGN

Ideating With AI

We began ideation by using AI to rapidly generate a broad range of dashboard concepts based on our research findings and design requirements. These concepts were never treated as final solutions. Instead, they served as conversation starters that helped the team explore layouts and align on direction before proceeding to wireframing.

Click to Zoom

DESIGN

Design System & Accessibility

Rather than invent a new visual language, my team made a deliberate call to work within the established design documentation outlined by the Center for Civic Design and Michigan's Digital Guidelines. These established tokens and components are already trusted by government stakeholders and built to work within Power BI constraints. Starting from that foundation meant our design decisions could focus on information architecture and interaction patterns rather than fighting institutional expectations or platform limitations.

Every new component was evaluated for feasibility within Power BI and tested against WCAG 2.1 AA standards using WAVE and axe DevTools. All critical accessibility issues were resolved before handoff, while remaining refinements were documented for implementation.

DESIGN

Key Decisions

The redesigned dashboard was built around three core interactions. Each one targets a specific pain point our research has uncovered.

DESIGN

Key Decisions

The redesigned dashboard was built around three core interactions. Each one targets a specific pain point our research has uncovered.

Misinformation
Prebunking

Cascading
Search Filters

Data Sharing &

Exportation

Click to Jump to Each Section

DESIGN

Misinformation Prebunking

Early in our research, journalists and clerks flagged the same concern. People weren't just confused by the data, they were arriving with suspicions already formed. A number that looked unusual without context wasn't just hard to read, it was easy to misread deliberately. That reframed the prebunking components from a nice-to-have to a core feature.

DESIGN

Misinformation Prebunking

Early in our research, journalists and clerks flagged the same concern. People weren't just confused by the data, they were arriving with suspicions already formed. A number that looked unusual without context wasn't just hard to read, it was easy to misread deliberately. That reframed the prebunking components from a nice-to-have to a core feature.

DESIGN

Misinformation Prebunking

Early in our research, journalists and clerks flagged the same concern. People weren't just confused by the data, they were arriving with suspicions already formed. A number that looked unusual without context wasn't just hard to read, it was easy to misread deliberately. That reframed the prebunking components from a nice-to-have to a core feature.

Original Design

A single disclaimer banner at the top of every page: "These results are unofficial until certified by the Board of State Canvassers."

Why it fell short: A blanket disclaimer at the top of the page taught users to tune it out. It also treated all data as equally uncertain, which isn't accurate — certification status varies by race and jurisdiction.

A single disclaimer banner at the top of every page: "These results are unofficial until certified by the Board of State Canvassers."

Why it fell short: A blanket disclaimer at the top of the page taught users to tune it out. It also treated all data as equally uncertain, which isn't accurate — certification status varies by race and jurisdiction.

Initial Iteration

Certification status moved next to each race header, with color-coded tags (unofficial / certified). Info icon opened a modal explaining the certification process.

Why we kept going: Testing showed users clicked the info icon after they had already interpreted the data — too late to change their read. The context needed to arrive with the numbers, not behind them.

Certification status moved next to each race header, with color-coded tags (unofficial / certified). Info icon opened a modal explaining the certification process.

Why we kept going: Testing showed users clicked the info icon after they had already interpreted the data — too late to change their read. The context needed to arrive with the numbers, not behind them.

Final Design

Contextual anchors built directly into each data view. Certification status appears next to the numbers it modifies. Info icons let users learn why variance happens — before they have to ask.

What Changed

The lesson across three iterations: context has to arrive at the moment of interpretation, not before it (banner) or after it (modal). By the third pass, prebunking wasn't a warning bolted on — it was baked into how the data itself was presented.

Contextual anchors built directly into each data view. Certification status appears next to the numbers it modifies. Info icons let users learn why variance happens — before they have to ask.

What Changed

The lesson across three iterations: context has to arrive at the moment of interpretation, not before it (banner) or after it (modal). By the third pass, prebunking wasn't a warning bolted on — it was baked into how the data itself was presented.

(This is the conclusion to the above items)

Misinformation Prebunking

(This is the conclusion to the above items)

(This is the conclusion to the above items)

DESIGN

Cascading Search

The original site had no search at all. Our first instinct was a standard global search bar, but usability testing revealed that voters, clerks, and journalists were asking fundamentally different questions of the same data. A single search couldn't serve all three without overwhelming each one.

DESIGN

Cascading Search

The original site had no search at all. Our first instinct was a standard global search bar, but usability testing revealed that voters, clerks, and journalists were asking fundamentally different questions of the same data. A single search couldn't serve all three without overwhelming each one.

DESIGN

Cascading Search

The original site had no search at all. Our first instinct was a standard global search bar, but usability testing revealed that voters, clerks, and journalists were asking fundamentally different questions of the same data. A single search couldn't serve all three without overwhelming each one.

Original Design

A single global search bar with autocomplete: "Search elections, races, or counties..."

Why it fell short: Autocomplete surfaced too much at once. Users saw hundreds of matching results and didn't know which to pick. Journalists especially wanted structured filters, not a lucky-guess text box.

A single global search bar with autocomplete: "Search elections, races, or counties..."

Why it fell short: Autocomplete surfaced too much at once. Users saw hundreds of matching results and didn't know which to pick. Journalists especially wanted structured filters, not a lucky-guess text box.

Initial Iteration

Three-filter approach: Election Year → Race → Jurisdiction. Cleaner than the search bar, but testing revealed users still didn't know where to start.

Why we kept going: Users kept selecting filters in different orders and hitting dead ends when combinations didn't exist. We needed the filters themselves to guide the sequence.

Three-filter approach: Election Year → Race → Jurisdiction. Cleaner than the search bar, but testing revealed users still didn't know where to start.

Why we kept going: Users kept selecting filters in different orders and hitting dead ends when combinations didn't exist. We needed the filters themselves to guide the sequence.

Final Design

Five cascading filters that narrow progressively: Year → Election Type → County → City/Township → Office. Each dropdown updates based on the previous selection, so users can't select an invalid combination.

What Changed

The lesson across three iterations: for a data set this granular, users needed the interface to guide the question itself, not just answer it. Cascading filters replaced open-ended search with a structured path — one that Power BI could support natively through cascading slicers.

Five cascading filters that narrow progressively: Year → Election Type → County → City/Township → Office. Each dropdown updates based on the previous selection, so users can't select an invalid combination.

What Changed

The lesson across three iterations: for a data set this granular, users needed the interface to guide the question itself, not just answer it. Cascading filters replaced open-ended search with a structured path — one that Power BI could support natively through cascading slicers.

(This is the conclusion to the above items)

Misinformation Prebunking

(This is the conclusion to the above items)

(This is the conclusion to the above items)

DESIGN

Data Sharing & Exportation

The original site had a single download link that dumped every race in an election into one massive TXT file. From presidential, to senate, to every district representative. Data would have to be scraped line by line. Users needed data scoped to what they were actually researching.

DESIGN

Cascading Search

The original site had a single download link that dumped every race in an election into one massive TXT file. From presidential, to senate, to every district representative. Data would have to be scraped line by line. Users needed data scoped to what they were actually researching.

DESIGN

Data Sharing & Exportation

The original site had a single download link that dumped every race in an election into one massive TXT file. From presidential, to senate, to every district representative. Data would have to be scraped line by line. Users needed data scoped to what they were actually researching.

Original Design

A single "Download" button that exported the currently viewed race as a CSV.

Why it fell short: Better than the original TXT dump, but journalist interviews revealed different downstream tools. Some wanted Excel with formatting preserved. Some wanted JSON for feeding into analysis pipelines. One format left users converting files themselves.

A single "Download" button that exported the currently viewed race as a CSV.

Why it fell short: Better than the original TXT dump, but journalist interviews revealed different downstream tools. Some wanted Excel with formatting preserved. Some wanted JSON for feeding into analysis pipelines. One format left users converting files themselves.

Initial Iteration

A download button that revealed three format options on click: CSV, Excel, JSON. Users could pick their format at export.

Why we kept going: Format selection worked, but users still didn't know what they'd get — full data set, filtered data, current race only? Testing surfaced hesitation before every download.

A download button that revealed three format options on click: CSV, Excel, JSON. Users could pick their format at export.

Why we kept going: Format selection worked, but users still didn't know what they'd get — full data set, filtered data, current race only? Testing surfaced hesitation before every download.

Final Design

A download modal with three format options and a scope preview: "You are downloading the [Race Name] results — [Year], [Jurisdiction]." Users see exactly what they're getting before they click.

What Changed

The lesson across three iterations: downloads aren't just about format — they're about trust. Users want to know exactly what data they're getting before it lands on their machine. Adding a scope preview turned "click and hope" into "click with confidence."

A download modal with three format options and a scope preview: "You are downloading the [Race Name] results — [Year], [Jurisdiction]." Users see exactly what they're getting before they click.

What Changed

The lesson across three iterations: downloads aren't just about format — they're about trust. Users want to know exactly what data they're getting before it lands on their machine. Adding a scope preview turned "click and hope" into "click with confidence."

(This is the conclusion to the above items)

Misinformation Prebunking

(This is the conclusion to the above items)

(This is the conclusion to the above items)

FINAL DESIGN

The New Dashboard


FINAL DESIGN

The New Dashboard


FINAL DESIGN

The New Dashboard


IMPACT

Measurable Improvements

Users found the new interface easier and faster overall. Perceived difficulty dropped across most tasks, most notably on Task 5 (−64%), and average task times fell (−52%).

On the other hand, time increased even as difficulty stayed around 1.0 during Tasks 1, 2. A closer look at session notes shows participants pausing to verify results and explore the interface, suggesting the UI invites engagement without adding effort.

IMPACT

Measurable Improvements

Users found the new interface easier and faster overall. Perceived difficulty dropped across most tasks, most notably on Task 5 (−64%), and average task times fell (−52%).

On the other hand, time increased even as difficulty stayed around 1.0 during Tasks 1, 2. A closer look at session notes shows participants pausing to verify results and explore the interface, suggesting the UI invites engagement without adding effort.

52%

Average Task Time

43%

Average Difficulty

52%

Hardest Task Time

64%

Hardest Task Difficulty

REFLECTION

Impact of Civic Technology

Civic UX is a different animal. The usual goals such as reducing clicks, smoothing flows, cleaning up hierarchy, all apply. But the stakes of misunderstanding are much higher. A confusing label on a shopping site costs a conversion. A confusing label on an official election dashboard can feed a conspiracy theory. That reframed how I thought about every design choice I made.

It also changed how I think about who the design serves. The end user isn't always the one in front of the screen. Sometimes it is the person whose election result appears on it, or the journalist writing about it at midnight, or the official who has to defend it publicly. Designing for that full chain of consequence is the standard I want to keep working at.

REFLECTION

Impact of Civic Technology

Civic UX is a different animal. The usual goals such as reducing clicks, smoothing flows, cleaning up hierarchy, all apply. But the stakes of misunderstanding are much higher. A confusing label on a shopping site costs a conversion. A confusing label on an official election dashboard can feed a conspiracy theory. That reframed how I thought about every design choice I made.

It also changed how I think about who the design serves. The end user isn't always the one in front of the screen. Sometimes it is the person whose election result appears on it, or the journalist writing about it at midnight, or the official who has to defend it publicly. Designing for that full chain of consequence is the standard I want to keep working at.

REFLECTION

Impact of Civic Technology

Civic UX is a different animal. The usual goals such as reducing clicks, smoothing flows, cleaning up hierarchy, all apply. But the stakes of misunderstanding are much higher. A confusing label on a shopping site costs a conversion. A confusing label on an official election dashboard can feed a conspiracy theory. That reframed how I thought about every design choice I made.

It also changed how I think about who the design serves. The end user isn't always the one in front of the screen. Sometimes it is the person whose election result appears on it, or the journalist writing about it at midnight, or the official who has to defend it publicly. Designing for that full chain of consequence is the standard I want to keep working at.

View Prototype

Other Projects...

View All Projects

BACK

Overview

Create a free website with Framer, the website builder loved by startups, designers and agencies.