Pal is a mobile companion for people with Parkinson's, connected to a view for their caregiver and one for their neurologist. Each morning the patient does three short exercises (voice, writing, mobility) and can tell Pal how the night went in her own words. Over time, those readings fill in the months between clinic visits.
Parkinson's changes from day to day, and within a single day. A neurologist sees the patient every few months and relies mostly on what she remembers. Tracking can fill that gap, but only if it fits the life of someone with a tremor and daily fatigue. A task that feels like a test gets dropped within a week.
We had input on the needs of patients, caregivers and clinicians, but no direct access to users. So we worked from explicit assumptions, listed at the end. The screens here follow one patient, Anna, through a good day and a bad one.
The obvious approach is to score every reading against a clinical scale or a healthy baseline. For most patients that means landing below the line every day, forever. It also isn’t what the clinician needs most: they want to know how this particular person is changing.
Every reading in Pal is compared with the patient’s own last seven days. That range sets the status. The sentence on screen, though, always points to a day she can remember: “Two seconds longer than Monday”, never “below expected”. In the app, her range is simply “your usual”. The home screen opens with one word for the day (Steady, Watching or Needs care), and the product name sits small in the corner.
A tracing exercise normally asks for one continuous line. For someone with a tremor, that means failing at the very thing being measured. In Pal she can lift her finger whenever she likes, and Pal stitches the strokes together. The voice ring never turns red and never completes, because there is no target to miss. However long she speaks counts as done.
The trust question comes up at the moment of recording, so the answer is on that screen, not in settings: the audio stays on the phone, and only the number travels.
Coming back to the concept for this case study, I found gaps and designed them afterwards.
The original set only showed good days, and a product for a progressive condition has to hold up when the numbers go down. On a bad day Pal states the change, puts it in context (“runs like this are part of your pattern”), offers a shorter set, and leaves the rest for tomorrow. No cheering, no blame, no alarm.
Anna's day reaches two other people: Dr. Ferrari, her neurologist, and Paolo, her caregiver. Two screens explain, in her language, what each of them actually receives. Every item is labelled measured or self-reported, and every item has a switch: her exact words, her recordings, or just her daily numbers.
The audience is 60+ and many users have a tremor. The original screens set meaning in 10–11px labels, at contrast as low as 2.9:1. I brought every meaningful label to at least 13px and 8.8:1, turned seven text links into buttons at least 48pt tall with 20pt of clearance from the primary action, and wrote one rule for completion verbs.
There is nothing to compare against yet, so Pal names the gap and says when it ends, instead of inventing a baseline.
Claude Design was part of every phase: brainstorming directions, prototyping flows fast enough to compare alternatives, and later the accessibility audit and the missing states.
Pal is a concept and was never tested with patients. The assumptions I'd check first:
And one question the design doesn't answer yet: when does a trend stop being a pattern and become a signal, and who tells Anna? On a bad day she's reassured, while the clinician is reading a long run of gradual decline. I think that threshold should be set clinically, not by the design, but the product still needs a moment for it.