Multifactor Authentication in University Settings

Overview

Our team was sponsored by Duo Security to conduct a usability study of their authentication app across users at the University of Washington.

We conducted a survey and recruited 6 participants to take part in a task-based moderated testing session. We utilized time-on-task andpost-task Likert scale metrics, coupled with a think-aloud protocol to gather qualitative and quantitative data.

Timeline

January - March 2024

Methods

Surveys, Moderated Testing

Role

UX Researcher

Impact & Outcomes

6

User Testing Sessions

23

Survey Respondents

3

Implementation Priorities

Key Findings

Clear Mental Models

Users can generally anticipate what they need to do with Duo for a typical Canvas log-in process.

2FA Misconceptions

A segment of users don't understand the purpose of 2FA, and express frustration about the additional step needed to log in.

Setup Friction

Linking a new platform with Duo's 2FA was a journey with unclear steps due to context switching between platforms.

Discovery & Planning

We began with stakeholder interviews across different roles within the Cisco Duo team to align our research objectives with their business goals. Through these discussions, we identified key areas of investigation and developed our research strategy.

🎯

Research Objectives

Primary Goals
  • Understand mental models and authentication workflows
  • Identify pain points in setup process
  • Examine user behavior differences
πŸ“‹

Methodology Design

Research Approach
  • Mixed-methods: Surveys (n=23) + Usability Testing (n=6)
  • Active UW students and staff participants
  • Focus on Duo authentication workflows

Stakeholder Insights

πŸ‘₯
User Experience

Focus on reducing friction in authentication flows

πŸ”’
Security

Balance security with user convenience

πŸ“±
Platform

Cross-platform consistency and integration

User Research

Our research execution phase consisted of two main components: a broad survey to gather quantitative data and in-depth usability testing sessions to observe actual user behavior.

πŸ“Š

Survey Design & Distribution

Distribution Channels
  • UW student channels
  • Staff communication networks
  • Department mailing lists
Key Metrics
  • 23 responses collected
  • Two-week collection period
  • Focus on authentication patterns
πŸ”

Usability Testing Sessions

Session Structure
  • 6 moderated sessions
  • Diverse user profiles
  • Think-aloud protocol
Data Collection
  • Mental model exploration
  • Service setup scenarios
  • Post-task interviews

Research Timeline

1
Survey Launch

Week 1

2
Data Collection

Week 2

3
Testing Sessions

Week 3

Survey Results

Respondent Demographics

  • 23 Respondents
  • 12 Male, 12 Female, 1 GNC
  • 82.6% Students, 17.4% Employee/Staff
  • 69.6% iPhone Users, 26.1% Android Users, 4.3% Multi-Phone Users

Usage Insights

  • 87% interact with Duo at least once a week
  • 39.1% unaware they could link personal accounts with Duo 2FA
  • 82.6% have fewer than 3 linked accounts with Duo

Usability Study Results

Likert scale ratings from 1-7, where 1 represents negative attributes and 7 represents positive attributes.

Task 1

100% Success

Description

Think-aloud process of logging into Canvas with Duo (no device)

User Sentiment

"Intuitive, matched my mental model."
* Verbal walkthrough only - no metrics collected

Task 2

100% Success

Description

Actual login to Canvas using Duo Mobile

Time on Task

1.58 mins

Perceived As

Very Easy Very Short

User Sentiment

"Easy, quick, no issues."

Task 3

33% Success

Description

Add a third-party app (e.g., Gmail) to Duo for 2FA

Time on Task

9.8 mins

Perceived As

Hard Long

User Sentiment

"Frustrating," "Impossible," "Confusing."

Task 3 Failure Analysis

4/6

participants failed to complete the task

6-18

minute completion time range, indicating high variability in user struggles

Reflections

Our study revealed that while Duo's 2FA system is generally effective, there's a significant opportunity to improve the onboarding experience and reduce friction during third-party service integration.

We found that users' mental models of authentication were consistent with their actual actions during the task of authentication. Successfulness of linking new accounts with Duo's authentication service varied based on their technical background, suggesting a need for more adaptive and contextual guidance during complex setup processes.

Moving forward, we recommend focusing on streamlining the third-party integration process, increasing instructional clarity, and building messaging that educates users about the advantages of two-factor authentication solutions and promotes trust.

Research Artifacts

Github Linkedin Email

Developed with Svelte.

Β©Wilson Chen 2025