Role: My role during this study was as a User Experience Researcher I embedded within our mobile team. The mobile team was the team that was developing our mobile interface for our hardware insulin pump.
Study Goal
Like many other experiences, medical devices are becoming increasingly linked with our smartphones. In this research, I was working on Tandem’s mobile app, which connected to a paired insulin pump via Bluetooth. The goal of this study was to understand our user’s expectations for how insulin pump menu navigation could be displayed on a mobile app and discover any existing mental models they may have for that type of system.
Challenges
The main challenge of this study was scoping the extent of menu items to focus on. This study would need to ensure both high-level menu headers were understood, as well as the action items contained within each menu item. Like with many medical devices, there were many settings and options to control the paired device, and while they may have a natural place to be on the medical device itself, how it might appear on a mobile app was uncertain.
Process
- Reviewed feedback from previous usability tests, which initially identified user problems with menu navigation.
- Selected a card sort methodology and met with product and UX teams to understand any limitations to menu structure within the mobile app and compile a list of menu items to be included in testing.
- Contacted recruitment vendor to establish a timeline for participant recruitment.
- Drafted a moderator script and built out the card sort activity on Optimal Sort.
- Set up a dry run with stakeholders to go over the execution of the research design and provide an opportunity for feedback.
- Executed the card sort over 1 week, running 4 participants per day.
- Compiled the results of the card sort into a PowerPoint which highlighted participant grouping trends and my recommendations.
- Worked with UX and product teams to develop menu structures based on card sort results, which would be used in follow-up A&B testing.
Methodology
This research study used an open card sort methodology. I chose to use an open card sort for this study due to a lack of understanding of how our users expected the insulin pump menu structure to be displayed on a mobile app. During the card sort, participants used the tool Optimal Sort to group a total of 36 menu items into categories that they created. Data collected: Quantitative tallies of groupings and qualitative comments and feedback made during the activity. The study included a total of 18 users across four distinct user groups defined by a previous clinical literature review. I requested our recruitment vendor to try and recruit older participants for the adult group as they had proven to have the most difficulty with our mobile app in the past. The users recruited are represented below:
| User Group | Count |
| Adults (53 to 70 years old) | n=5 |
| Adolescents managing their own therapy (15 to 19 years old) | n=5 |
| Parent-child dyads with shared responsibility (children 9 to 13 years old) | n=5 |
| Parents responsible for child’s therapy (children all 8 years old) | n=5 |
Key Findings
The feedback provided by participants showed support for 4 very strong and consistent categories across the board.
1. Alerts: includes alerts, reminders, push notifications and sounds.
2. CGM: includes starting/stopping sensor, setting transmitter ID, calibrating, and CGM info.
3. Pump/Insulin Settings: includes all insulin profile settings, likely includes profile activation, and potentially includes setting pump time and date.
4. App: includes help resources, information about the app, t:connect account management, includes pairing/unpairing phone and pump.
In total, users created an average of 6 groups, with the lowest being 4 and the highest being 12. I also opted to create a visual representation of user groups to give stakeholders a better understanding of what was observed in the study.
Outcome
The results from this study supported the UX team in the development of new designs for the app’s menu structure. These designs were then narrowed down in follow-up A&B testing, which helped improve user satisfaction scores by 34% upon the next formative usability test. The improved satisfaction scores were a result of the inclusion of helper text in the high-level menu headers as well as a reordering of menus, which followed the recommended information architecture that was proposed (found below).
