This was really two systems. The app, where a worried person types what they feel. And MediBase behind it, where the medical experts set the rules for every condition the app is allowed to check.
What was missing
"I couldn't find the words for what I was feeling."
People don't describe symptoms in neat terms. They write long, they leave out the specific part, they guess the spelling. And whatever went into Healthily's one free-text box shaped what came out. Fifteen interviews and eight usability tests found the same pattern. Their own words never made it in cleanly.
So I added a step after the language processing. It reads what someone types and shows back the symptoms it heard. Then autocomplete lets people sharpen that list, adding what is missing and taking out what is wrong. Each symptom carries its everyday names too, so "tight chest" lands on the right entry with no perfect spelling required. Symptom-description accuracy rose by 23%.
Pain points. Medical terminology, spelling errors misread, no guidance during entry.
What changed. Autocomplete, “also known as” terms, visual confirmation, editable symptoms.
However a user puts it
Symptom input · however a patient types it
Recognition & triage mapped to one outcome- Self-care at home
- Pharmacy
- See a GP
- Urgent care · 999emergency
The knowledge base behind it
Under the search box sat MediBase. I structured it and led it: every condition, its symptoms, its warning signs and the rules for what the checker could safely ask. Every mapping was tested and approved by an expert before it went live, across a library of thousands. The medical calls were never mine. I built the system the experts used to make them.
Why the smaller version lost
Adding a step to a medical flow is not free. Every extra screen is somewhere to give up, and someone who quits halfway never gets told to see a GP.
So rather than guess, I put two versions in front of users. A was autocomplete alone, the smaller build. B showed the everyday names under each suggestion, so “runny nose” could be recognised without knowing the medical term, plus a way out when nothing fitted.
A tested worse. Autocomplete hands you a list, but no way to check the medical term is the thing you meant. People wanted their own words back before committing.
The drop-off never came. Satisfaction rose from 62% to 67%, and the 23% accuracy gain held.
Three products, one source of truth
Across the website, the web app and the app, this had to feel like one thing. The Smart Symptom Checker was mine across all three, its visual language and its wording, so a symptom looked and read the same at every touchpoint. That took six teams behind one source of truth, design, legal, the experts and marketing among them, and holding them there was my job. Behind it sat MediBase, the backend I led, where the experts kept every condition updated and customised.
The homepage, top to bottom.
The home for Dot, the symptom checker.
Dot greets you and offers a route in.
Your day and your trackers.
also known as: chest tightness
lay terms: chest pressure, tight chest
✓ Ratified · clinician-signed
- 1Publish + red-flag gatesNothing reaches a patient un-reviewed.
- 2Symptoms keyed to UMLS CUIs, weightedWeighted scores across coded symptoms decide the outcome.
- 3Inclusion / exclusion rule engineThe logic that decides each outcome.
What it changed
Symptom-description accuracy rose 23%. Completion rates went up, and over time the app's star rating climbed as onboarding got easier. As design lead of the symptom-checker squad, a junior designer and a brand designer with me, I carried the work end to end, research through rebuild.
Reflection
The problem showed up at the end of the flow: consultations ending with no outcome, low completion, low ratings. The interviews found the cause at the start. People could use the app fine, they just couldn't get their own words into it. Fixing the input lifted both the accuracy and how people rated onboarding. The interviews found the fix. The redesign followed what they said.