TL;DR
- What was at stake
- The company needed to grow past friends and early adopters to a real user base before its runway ran out.
- What changed
- The founders added employer and union channels alongside direct-to-consumer, and started re-engagement emails for trial users who never converted. Outreach to advocacy groups such as NAACP was underway when the engagement ended.
- My role
- Solo researcher and strategist for the founders, from study design and recruiting through sessions, behavioral analytics, and recommendations
- Timeline
- About 3.5 months: 6 weeks of research through the January readout, then advisory through March
Users, up from fewer than 20, mostly friends and people paid to sign up. Past the founders’ 10,000-user target.
Six-figure service agreement, which came directly from the employer and union channel.
Employer and union channels alongside direct-to-consumer, and re-engagement emails for trial users who never subscribed.
Partway through building personas, I realized sharper targeting wouldn’t help. People either didn’t trust the product or didn’t think it was for them.
Read the full case study
Meet people through someone they already trust
The Trust-Transfer Model introduces the product through an institution people already trust, like their employer or union.
Recommendation
Partner with self-funded employers, stop-loss carriers, unions, and other trusted groups to offer the product as a benefit. Start where the value is easy for the employer to see: burnout among healthcare workers, time lost by professionals caring for aging parents, and absenteeism in manufacturing.
Channel
Seek endorsements from groups like the Alzheimer’s Association, caregiver alliances, AARP, and NAACP health initiatives, so people who find the product on their own see a name they recognize.
Before Scale
Test the partner approach with a 30-day pilot before shifting the whole go-to-market plan, with a clear point to stop if it didn’t reduce the friction I saw in research.
Quick Win
Behavioral analytics showed no follow-up with people who started a free trial and didn’t subscribe. I recommended a short re-engagement email sequence for them, since the cost of reaching them had already been spent.
The partner channel was the bigger bet, so I also recommended a 30-day pilot with a clear point to stop before the founders moved their whole go-to-market.
The founders saw a product problem
Adoption is a product or feature problem.
The interface had a few rough spots, but nothing that explained the drop-off. People were leaving at the entry point.
People who backed out of participating had something to add, even without trying the product.
What would have changed my mindIf the people who backed out had given usability reasons, like setup taking too long, I’d have kept the focus on the product.
The reasons people gave for backing out became the trust finding.
A product that worked and a growth problem
The client was an early-stage AI health platform built to work like a personal medical command center: one place for your health history, appointment prep, and a check on possible diagnostic errors. The technology worked and had early clinical backing. Getting people to sign up and stay was the problem.
I was brought in to run usability research and find out where people were dropping off. I worked directly with the founders, with no PM or design team in between, and presented findings that challenged the go-to-market plan they had built.
Following the people who said no
I recruited and screened everyone myself, across a range of ages, health engagement, and comfort with technology, and ran the sessions over Zoom.
The screener stayed open through the break, so sessions could start the first week of January.
Moved the question from usability to who people thought the product was for.
Turned the findings into a channel recommendation.
Eight sessions were enough to find the pattern. The 11 people who backed out added evidence for the trust finding.
A pipeline that saved 6 to 7 working days
I set up the analysis as a chain, with AI doing the first pass at each stage and me checking its work before the next one started. A short script I wrote handled the file conversion, so transcripts went from recording to coding in a day.
- AITranscribeSessions transcribed automatically.
- Script I wroteConvertI wrote a script to convert transcripts to .vtt, so they loaded into Dovetail the same day.3–4 days saved
- MeVerifyEvery transcript checked against its recording, since the automated ones had errors.
- AIClusterDovetail’s AI grouped themes across sessions as a first pass.~2 days saved
- MeCodeI did the coding from there, starting from the AI’s groups.
- AIDraftNotion AI drafted slide copy from my theme notes.~1 day saved
- MeTriangulateSaved time went into checking what people said against the behavioral data.
People liked it. Most decided it wasn’t for them.
Behavioral analytics and interviews indicated the same thing. The bigger findings were about whether people saw the product as something for them.
All 8 accepted the concept. 7 still saw the product as meant for someone else.
Refused over privacy and trust.
Most named someone else it would help: a parent, a grandparent, a friend with a chronic illness. Some said they would use it if their doctor told them to. That’s optimism bias, the belief that health risks apply more to other people, and I named it the “Not for Me” Paradox so the founders had a handle for it.
The people who backed out had already heard the product was HIPAA and SOC 2 compliant. For them, feeling safe was something the product hadn’t earned yet, and a badge on the page couldn’t supply it.
The platform never said whether a doctor was right or wrong. It listed possible diagnoses, each labeled with a level of likelihood, and the doctor’s diagnosis might or might not be on the list. More than half of the participants were skeptical of anything out of the ordinary, and a high-confidence label on an unusual diagnosis made them question it more. They wanted someone they already trusted to back it up, which became another piece of the logic behind the trust-transfer recommendation.
“I totally get why someone with a chronic illness or like an elderly parent would need this. It would actually be really great for my nana, but personally, I’m super healthy so I don’t need it.”
“If my primary told me, ‘Hey, download this app before our next visit,’ I would probably do it. But just finding it on my own? I’d assume it’s for people with way more complex medical histories.”
“I saw the page says it’s HIPAA compliant, but nowadays everyone says they’re secure until a data breach happens. I’m just uncomfortable uploading my full medical records to a startup I’ve never heard of.”
Impact
This was the first time I delivered something other than what a client asked for, knowing it might not be welcome. For the founders, it replaced guesswork about who to target with a clear reason people weren’t signing up.
Hindsight
The research pointed to the entry point as the problem, but the sessions covered the whole product. A study built only around first-time activation would have shown exactly where trust broke down.The founders would have had a specific screen to fix alongside the channel change.
The founders’ changes were about channels, not the interface, so there was nothing new to bring back into testing. I would push for one onboarding change alongside the strategy work, so a follow-up study could measure the result.The founders would have had a before-and-after number to bring to partners, not just user growth.
Caregivers came up during analysis as a strong segment, but I hadn’t recruited them on purpose. Building them into the screener from day one would have given that group its own findings and likely sped up the partner recommendation.The advocacy outreach could have started weeks sooner, with findings behind it.
With two more weeks
Recruit caregivers directly and give them their own sessions.
Employers decide through HR and benefits teams, so the partner pitch needs testing with them too.