115 laps at Thunderhill
A year of my own telemetry, against a theoretical-best floor.
Burnout as a heat map
A study of medical student mental health, measured on two validated instruments and read as density instead of as points.
Built for the Cleartelligence Chart Champ competition, where it placed fourth. The top five entries were shown at the live event. Mine ran as a prerecorded walkthrough, played to the audience at Boston University in one pass. Designing for that is a different job from designing for a reader who can lean in.
An entry for Chart Champ 2023
Cleartelligence ran its eighth annual Chart Champ contest on 4 October 2023 at Boston University. The theme that year was higher education, dashboards were due 22 September, and the five highest-placed entries were presented live at the event.
The subject I picked under that theme is medical student mental health: survey scores for depression, burnout and exhaustion, set against curriculum year and weekly study hours. The measures correspond to a public dataset whose fields include cesd, stud_h, health and the MBI subscales.
Two panels, because two questions
The data offered more than two questions and the canvas offered room for two. Depression severity against curriculum year went left. Weekly study hours, the variable a student can actually change, went right, carrying three stacked measures over one shared x axis.
Density marks were the second decision. A few hundred students plotted as points is a smear, and a binned heat map introduces cell edges that read as structure the data does not have. Tableau's density marks are kernel-smoothed, so repeated responses at the same coordinates accumulate into continuous regions with a readable center.
Density marks across two panels
The canvas is 1,366 by 795. On the left, severity of depression symptoms runs 0 to 60 against curriculum year 1 to 6, on a grey-to-teal ramp. On the right, three strips stack over a shared x axis of hours of study per week, 0 to 75: academic burnout at 0 to 40, emotional exhaustion at 0 to 35, and mental health at 1 to 6, all on a grey-to-orange-to-red ramp.
A pale grey silhouette of a head in profile occupies the right quarter of the canvas and carries no data. Axis titles and tick labels are the only text on the dashboard.
Impact
The dashboard placed fourth overall. Because the top five were shown at the live event, the work did not end at submission. I recorded a walkthrough of it that played to the audience at Boston University.
That changes what a chart has to do. A dashboard read alone gets as long as the reader wants. A dashboard shown in one recorded pass gets the speaker's pace, from the back of the room. The two-panel split is what survived that constraint. Depression severity against curriculum year on the left, three strips over weekly study hours on the right, and the whole argument reduces to a shape people can see before I finish the sentence describing it: the load does not taper as students advance.
Density marks earned their place for the same reason. A scatter of a few hundred students is a smear from twenty feet away. Kernel-smoothed density turns repeated responses into a region with a readable center, so the pattern holds at presentation distance instead of dissolving into ink.
What I would change is the thing the room could have asked about and did not. The published dashboard carries no title, legend, units note or source line, because the recorded walkthrough was supplying all of that. Once it was back on Tableau Public without the recording, that context was gone. A chart that has to travel without its author needs to answer those questions on its own face.