Microvisualizations on Fitness Trackers

Visualization Research Intern

February 2019 - July 2019

Inria, France



With the guidance of my supervisors, I conducted 10 semi-structured interviews with regular wearers of fitness trackers (e.g. Apple Watch, Fitbit, Garmin, Xiaomi) to find out how often they use their tracker, what types of data they track and which insights do they pay more attention to.

Then, without looking at their fitness tracker, I asked them to roughly sketch on a sheet of paper what types of data they remembered seeing on their wearable and how they are visualized. Then, I asked them how their use of the associated phone or web app differed from their use of the fitness tracker.

Figure 1: One of the participants drew the current visualizations she remembered seeing on her tracker on the top half. In the bottom half, she gave suggestions on how these visualizations could be improved so that she can gain more insights.

In the final part of the interview, I asked the participants to do a card sorting exercise in which we gave participants 33 different smartwatch-sized data visualization cards. Each participant organized these cards into self-defined groups. Then, I generated a similarity matrix and a hierarchical clustering dendrogram to analyse the groups.

Figure 2: In the card sorting exercise, we gave these 33 visualizations showing fitness data. They were inspired from the visualizations on current trackers.

Figure 3: One of the participants organising the cards into different self-defined groups.



To learn more about visualization use and preferences for sleep data, we designed and deployed an online questionnaire on Google Forms. The questionnaire included questions about fitness tracker details, tracking behaviour, and preferences for seeing sleep data. We ran these questionnaires twice (second time with some updates) because we wanted to get a few more insights the second time. We received 146 and 108 responses in the two questionnaires respectively.

The first section had questions about the type of tracker, usage period and frequency. In the second section, there were questions specific to sleep data like “When do you check your sleep data on your tracker?” (to understand the context) and “Regardless of what your tracker can actually show right now, what kind of sleep data would you like to see directly on the tracker?” (to understand their preferences). In the final section, respondents gave their preferences for sleep visualizations representing sleep duration and phases data in various granularities: previous night, last week overview, monthly overview and social comparison. These visualizations were designed by me and inspired by designs from commercial trackers.

Figure 4: Preferences of 108 respondents for different types of sleep data. Orange borders highlight the most preferred designs.

Figure 5: Responses to question: "Regardless of what your tracker can actually show right now, what kind of sleep data would you like to see directly on the tracker?"

Figure 6: Responses to question: "When you check sleep data on your tracker, do you pay attention to the visualizations or the numbers?"


I am currently preparing to run an experiment on a smartwatch to test how screen size (smartwatch vs wristband) affects readability of sleep charts. I will ask 12 participants to do simple reading and comparison tasks. There has been a similar perception study glanceability study done recently on a smartwatch to test the accuracy of reading various charts and I would use the same methodology to test for glanceability of sleep visualizations.