A daily paradox.
The dietary decisions that most affect patients’ health are the ones they feel least equipped to make.
How might we reduce the daily cognitive burden of dietary management for diabetes patients, while keeping them meaningfully connected to their healthcare providers?
12 patients. 4 rounds. Real findings.
| Study setup | Who took part | Prototype | What we tested |
|---|---|---|---|
| Interviews | 12 patients · Type 1 & 2 · ages 35–74 | — | Daily routines, barriers, the doctor relationship |
| Expert review | 3 doctors + 1 nutritionist | — | Clinical accuracy of key decisions |
| Round 1 | 5 people · 4 of them aged 60+ | Paper | Navigation |
| Round 2 | 5 people | Mid-fi | Onboarding |
| Round 3 | 9 people | Prototype | Glucose card and GI signals |
| Round 4 · final | 9 people | Hi-fi | The full task flow |
| How I measured | Every round | — | Task success and time on task, then post-test questions and semi-structured interviews |
Michael, 35
Software Engineer · Prediabetes- Tech literacy: high
- Biggest barrier: meal planning time
I know what I should eat. I just can’t figure out what to actually cook every day.
Margaret, 65
Retired Teacher · Type 2 (10+ years)- Tech literacy: low
- Biggest barrier: app complexity
I find these apps confusing. The buttons are too small and there’s too much going on.
Ahmad, 74
Retired · Type 2- Tech literacy: very low
- Biggest barrier: doctor disconnect
My doctor sees me once a month. What I eat every day, he has no idea.
- Cognitive overload, not lack of motivation, was cited by every patient as the main reason dietary plans fail.
- Users 60+ abandoned every app tested because of interface complexity, not disengagement from their health.
- The doctor relationship was entirely absent from every competing product in the category.
No existing app combines AI meal planning, GI awareness, doctor integration and accessibility for 60+ users in a single product.
Four choices that shaped the design.
Accessibility-first navigation over feature density
A 3-item bottom nav with icon + label pairs and 60px touch targets — beyond the 48px WCAG minimum — instead of the 5-item hamburger every competitor uses.
Glycemic index as a visual signal, not a number
A 3-colour dot (green / amber / red) on every recipe instead of raw GI scores — most participants couldn’t interpret a raw number without explanation.
The doctor feedback loop, built into the core flow
A monthly PDF export with a dedicated doctor comment section — not real-time messaging, which doctors confirmed they couldn’t run as on-demand responders.
AI-generated meal planning over manual tracking
A personalised daily plan generated from each user’s diabetes type, age, glucose history and food preferences — refreshed every morning, without them having to think about it.
What testing revealed.
Before5-item hamburger nav
- 3 of 5 participants failed the navigation tasks
- Community tab was rarely discovered
- Older users abandoned within the first session
After3-item bottom nav
- Zero navigation failures in the final round
- All core functions one tap away
- Community content surfaced in the Home feed
Before7-screen onboarding
- Users felt interrogated before seeing any value
- High abandonment before reaching home
- Reported feeling overwhelmed
After3-screen onboarding
- Only age, diabetes type and weight upfront
- Other preferences collected progressively
- Completion improved significantly
BeforeRaw number display
- 7 of 9 couldn’t interpret mg/dL without clinical context
- Older users ignored the number entirely
- No link between reading and meal decisions
AfterPlain-language status
- In the final round, all 9 understood “On track” within 3 seconds
- Status linked directly to meal decisions
- Zero comprehension failures in final testing
The final designs.
Results.
All 9 final-round participants completed the full task flow without help, with zero navigation failures. All three doctors confirmed the export format would improve consultations.
Three honest takeaways.
What I learned
Accessibility should be the starting constraint, not a post-design check. Every major decision on this project was an accessibility decision in disguise.
What I’d do differently
Push the doctor integration earlier. The signals were there in the interviews, but it wasn’t prioritised until Round 2.
What surprised me
The competitive gap was larger than expected. No app addressed the doctor relationship at all — and that came from research, not assumption.