Cohorts: Putting Data in Context
Sparta Science is a platform that helps organizations understand how health and movement are connected. Physical therapy centers, sports teams, and the US military use Sparta Science's tools, like force plates and AI, to measure and track how people move. These tools help organizations improve people's health and physical performance.
Sparta's biggest clients needed more specific insights. Comparing people from different groups, like athletes and seniors, can give confusing results. The US Military and sports teams wanted to compare their users to similar groups instead of the entire diverse dataset. Comparing a college basketball player to a senior citizen in physical therapy is not helpful. By grouping similar people together, organizations can get a clearer picture of their performance.
Grouping data, or cohorts, emerged as a solution to this need. Sparta’s data scientists began categorizing user data based on characteristics such as age, weight, height, sports position, occupation, injury history, etc. This facilitated the formation of cohorts within the dataset, enabling more nuanced and relevant performance comparisons.
Problem
Due to HIPAA and other privacy concerns, designers had limited interaction with end-users. Instead, customer complaints were provided by sales and customer success teams, who regularly interacted with organization administrators. The customer success team shared these complaints with us, and I worked with a product manager to interpret these complaints into design requirements.
- There was a need to compare an individual's performance against a global population and a specific cohort.
- The user interface needed to allow easy switching between these comparisons while keeping everything clear and relevant.
- It was important to provide users with the ability to view a cohort's performance against the entire dataset.
Process
How might we visually contextualize performance metrics for cohorts?
I identified an existing dashboard screen and modified it to show comparisons to both cohorts and the entire Sparta dataset simultaneously with a dropdown menu. The current design featured data from the global population, visualized with bell curves, bar charts, and line graphs, all based on the global dataset.
Early designs used updated bell curves and other existing components to show different population sizes and metadata to show where the user fit in. However, usability testing on UserBob.com showed that users were confused by score changes and the metadata didn’t always make sense. Testing also showed users didn’t know how to read bell curves. The screen wasn’t working on a fundamental level. I shared these ideas with stakeholders and got their feedback. Based on this, I adjusted the designs and created new components to better meet user needs until I found a solution that worked well.
Solution
Further iterations led to the adoption of a modified box and whisker plot. Collaborating closely with the data science team, this approach ensured consistent performance context, enabling comparisons against both global and cohort populations while maintaining score integrity.
The final design features 3 main components:
- Light-colored background line representing the entire Sparta Science data set normalized to stretch from 0 to 100.
- The blue line in the foreground represents the normal distribution of this user’s cohort (e.g., a collegiate women’s basketball team)
- The white circle represents this user’s actual score. In this case, their Balance Sway.
Positive feedback from beta test clients confirmed that the new design exceeded their expectations. I wanted additional verification that the design worked, so I tested from a panel of non-biased clients to validated any assumptions before shipping. Testing showed that the design brought clarity and increased comprehension of cohort comparisons. I also made sure to consider accessibility by focusing on color contrast and screen reader support. The success of this approach has led to growing interest in updating the default screen to my new components.
