The research said I was solving the wrong problem.
Curalink is a peer-support community for cancer patients. I started it as a check-in tool. Three interviews later it was something else entirely, and so was everything I would say about it.
A note first: I'm a cancer survivor, which is why I went looking for this problem. Participant details are anonymised, and Cindy — the persona in this case study — is a composite.
APR Dr. Alvarez · 1:00 PMPrincess Margaret, Clinic B
I wrote the value proposition before I talked to anyone.
The original pitch was efficiency. Cancer treatment is heavily managed — appointments, protocols, check-ins — but the space between appointments is not, so patients lose track of symptoms. A digital check-in tool would save everyone time and energy. Desk research supported it: most patients record none of the symptoms they experience between sessions.
Then I interviewed three patients, and none of them wanted to talk about check-ins. All three described the same thing instead — people around them who cared but could not truly understand, and a wish to talk to someone who could. The problem was real. The message was aimed at the wrong one.
Three interviews is a small base. I treat it as direction, not proof — that holds for everything on this page.
A digital check-in tool that improves patient engagement and motivation during treatment
A place to talk to someone who has been where you are — emotional support and guidance from people who understand
How might we give patients the chance to connect with others who have gone through the same thing?
Cindy has a family and a career, but it is difficult for the people around her to truly understand what she is going through as she battles cancer. She longs for connection and community through this endeavor.
In a health community, trust is not legal copy. It's the retention mechanic.
Getting someone to sign up is the easy half. The hard half is the second post. Someone who writes about their diagnosis once and then cannot tell who read it, whether anyone is moderating, or how to take it back does not post again — they lurk, and a support community with lurkers is not a support community.
So the trust signals are not a policy page. They sit at the exact moment doubt appears, and they are the reason someone comes back. Four decisions carry most of that weight.
One audience setting — Everyone, a group, or only me — shown on the compose sheet and on the published post. Reach is never a surprise after the fact.
Every public surface names who is watching it: "looked after by staff and trained volunteer mentors, day and night." It's the first line of the feed, not a footer link.
Privacy is stated where the doubt arises. "Moderators never read your messages" sits inside the message thread, at the second a person hesitates to type.
Posting ends in proof rather than mystery: a preview exactly as the community will see it, sixty seconds of undo, and an edit that keeps existing replies attached.
Five moodboard rounds to land on three words.
Each round had a job. Real photography arrived at V2, colour at V3, the words at V4, and V5 balanced them. The three words that survived every round became the brief for everything after — the palette, the wordmark, and the tone of every line of copy in the product.
Blue for trust, green for progress, and lines reaching out to either side for movement. Connection was in the mark before it was in the product. Sans-serif throughout, picked for openness over ornament — the same reason the product's copy avoids clinical language.
It never launched, so there are no results. These are the three numbers I'd hold it to.
Signup counts would flatter this product and tell me nothing. A support community only works if people speak and get answered, so all three metrics measure that and nothing else.
The single honest test of whether the positioning landed. If someone joined for connection and never spoke, the promise on the way in was not the one they believed.
An unanswered post is worse than no community at all. This is the number I'd protect first, with moderators seeding replies long before it looked healthy on its own.
Patients post at 2am, and the interface promises "someone's usually up." Median hours is the number that says whether that line is true or marketing.
On writing the pitch too early
I had a value proposition, supporting research, and a plan before I had spoken to one person in the audience. The desk research was not wrong — it just answered a question nobody I met was asking. Three conversations moved the product further than months of secondary reading.
What I'd do next
Test the messaging with survivor communities rather than classmates, and bring a clinician and an experienced moderator into the safety model before any launch talk. Reddit and Facebook groups were where I found people — I'd treat them as the real acquisition channel and measure cost per member against each.