I designed dashboards for MyJourney, a cancer-care platform for clinicians, staff and patients. This is what I would have done differently, knowing what I know now.
PT-0938’s plan gives surgery 28 days. Today is day 36 of it. Before a whole clinic, the question for one person: where are they in treatment, and are they on schedule?
Try itSwitch between patients, or pick anyone from the clinic view in the next section.
| Before | After | |
|---|---|---|
| 1 | The stepper shows state, not time. A patient 8 days late looks identical to one who started today. | Planned against actual on one time scale. Stage lengths are drawn to scale and the delay shows where it happened. |
| 2 | The ring never says what 28 is. No target marker, and its orange means "current stage" one screen over. | The ring becomes a sentence. "Surgery is 8 days past its 28-day target." |
The dashboard clinicians opened first: how many patients are in each stage, and who is late. In the redesign every patient is a dot, so you can see where delays pile up without losing the people inside them.
Try itClick a number or a stage to filter. Click any dot, or a patient in the list, to open their timeline above.
| Stage | Active | 15+ days | 8-14 days | 1-7 days | On track |
|---|
| Before | After | |
|---|---|---|
| 1 | The donut says 158 patients; the stage cards add to 175. It only reconciles if three stages use totals and two use delayed counts. | One number leads. 106 of 175 patients are past target, with a 12-week trend. Every total reconciles. |
| 2 | Six rainbow hues for an ordered sequence. Amber means Radiation in one place and "Attention" in another. "10/12" is never labeled. | One dot per patient, one blue ramp. Position is lateness, nobody disappears into a bar, and the title states the finding. |
| 3 | Status with no rule. Why is 18 days "Attention" and 26 "Delayed"? Pale pills, tint on tint. | Bullet graphs. Days past target against visible bands and the stage median. |
The second product served a smaller, more specialized group. Before a patient’s treatment is set, their case goes to a multidisciplinary cancer conference (MCC): surgeons, medical and radiation oncologists, pathologists, radiologists and nurse navigators review it together and agree on a plan. Coordinators run these sessions every week for each cancer type, and program leads report on them by hospital and region.
The clinician dashboard answers “who needs me today?” The conference tool answers three different questions: which cases are ready to discuss, did the right specialists attend, and is volume keeping up?
For the coordinators who run each week’s sessions: which cases to discuss, who attended, and how volume is trending.
Try itPick a cancer type and watch every panel follow. Open a patient row. Hover the trend line.
| Patient | Age | Cancer type | Receptors | Stage | Reason for discussion |
|---|
| Before | After | |
|---|---|---|
| 1 | Five hues, one tangle. The lines cross constantly, so no trend reads. | One highlighted line. The other types stay gray for context, and the title states the change. |
| 2 | Filters that don't filter. The view says August 2021 to July 2022; the chart shows January to December 2022. The portal is titled Breast while its filter says Urologic. | One filter drives every panel. Counts, trend, attendance and table all follow the cancer type, and the counts add up to the total. |
| 3 | The same people again and again. The attendee list repeats three clinicians, with avatar colors that mean nothing. | Attendance by specialty against an 80% target. Each specialty once; the status line names who is below target. |
| 4 | Rows that change their story. The table repeats rows, mixes cancer type with receptor status, and a "Rectal" case becomes "Urologic" when opened. | Separate receptor column, expandable rows. Each row opens to the patient's pathway and the conference recommendation. |
| 5 | "75/96" with no label. Seventy-five of ninety-six what? | "28 of 34 cases discussed this week," with a meter and the percentage spelled out. |
For program leads comparing hospitals across a region: where conference sessions happen, and how many.
Try itChange the region. The title, bars, average and patient list all update.
| Patient | Reviewed | Hospital |
|---|
| Before | After | |
|---|---|---|
| 1 | "Insert graph here." The wireframe reserved a box for a chart before anyone decided what it should answer. | The question first. The title answers it: which hospital is busiest, and by how much. |
| 2 | The same numbers twice. The bar chart and the hospital table show identical counts. | One sorted chart with direct labels. The duplicate table is gone. |
| 3 | A quarter that isn't. The filter says January to March 2022; the patient list shows July and August. | Every row belongs to the selected quarter. |
| 4 | Hard-to-read labels, exposed IDs. "72.0" turned on its side, angled names, full health card numbers in a list. | Whole numbers, horizontal labels, masked IDs. |
One at a time, these look like small slips. Together, they are a system problem, and that is where I would fix them.
| Pattern | Care pathway tool | Conference tool | System fix |
|---|---|---|---|
| Unlabeled fractions | "10/12" on stage cards | "75/96" on session counts | A count component that won't render without a label for its denominator |
| Filters that don't filter | Status labels with no stated rule | Date range, quarter and cancer type disagree with the data shown | One filter state every panel reads, echoed in each title |
| The same data twice | Donut and cards disagree on the total | Bar chart repeats the table; attendees and rows repeat | One source per number, de-duplicated before it reaches the screen |
| Color without meaning | Six rainbow stage hues | Five line hues, random avatar colors | Color only when it encodes something: one ramp for lateness, one accent for focus |
Source: Herrington, Osman, Kastner et al., JMIR Cancer. Four staff participants across two clinics; the study evaluated the platform's clinical navigation tool, not this redesign.
The redesign was never tested; funding ended first. These are the targets I would set.
| Measure | How | Target |
|---|---|---|
| Find the longest-waiting patient | Timed task, original against redesign | Under 5 seconds |
| Name the busiest hospital in a region | One question per screen | 5 of 5 correct |
| Spot a delayed patient in grayscale | Same screens, color removed | 5 of 5 |
| Usability of the redesigned dashboards | System Usability Scale, same staff roles | At or above the live tool's 81.3 |
Method: charts built with Observable Plot. Bullet graphs after Stephen Few, sparkline after Edward Tufte, chart choices checked against the FT Visual Vocabulary, headline titles in the newsroom style. One blue ramp for lateness, checked for contrast and color-blind safety.