Keepsake

The problem

As of 2025, 63 million, or a quarter of adult US population provided ongoing care to a family member with a medical condition or disability, up 45% from a decade ago. On average, caregivers spend a staggering 27 hours a week on care. In the context of the shortage of medical personnel and aging population, this segment will only continue to grow.

On a macro level, home care is trending, and hospitals are increasingly creating task forces to implement hospital at home programs. I strongly believe that healthcare is increasingly moving to the home, and technology is needed to allow scale 1 to few to 1 to many.

Feature set

After a good amount of discovery, settled on the following features:

Task lists / logging. Manage daily tasks split by time of day, and log symptoms + observations. Hoped to promote a daily habit to learn more about users, build context, and provide valuable insights.

Appointment recording. Hit record and get summaries of appointments, generate task lists and prep sheets for upcoming appointments. Pretty straightforward and worked well with Apple's live captioning functionality.

Family sharing. Give access to family members, and provide easily sharable formats for appointment summaries, etc. We saw many of our users consisted of families with a primary caretaker with remaining family taking less responsibility, often separated by geography. Additionally, many primary users felt a strong sense of duty and had difficulty delegating responsibilities to other family, so we also hoped our platform could circumvent this as a shared source of truth that could make the experience more collaborative.

Community. Condition and topic specific forums for users to ask questions / receive advice. Inspired by a lot of facebook, reddit, and more online groups that were highly active and wanted to replicate that experience.

Guides. Curated duolingo-style curriculum, created in collaboration with our community partners like the Family Caregiver Alliance. Noticed that there were a lot of "unspoken" rules about caregiving learned through experience. Other companies have tried to address this problem with 24/7 professionals on call, care managers covered by medicare, and even certifying families as CNAs.

Monetization

Looking outward, two of payers’ and risk-bearing providers / ACO’s most costly challenges are member engagement and adherence.

Our vision was that by creating a great consumer app, we could create a structured data layer built upon voluntary tracking behaviors, and tailor recommendations to payer benefits and programs that are extremely difficult to enroll members in. This data could be extremely valuable for preventive care, cost containment, and budget allocation.

On discharge adherence, we were going after the large segment of non-adherers that fall-off due to overwhelming or misleading instructions. While this is a huge metric/problem gone after by the largest healthtech companies, overall improvements have been mostly incremental. We took the approach going user-side through families, hoping to leverage consumer psychology to unlock payer contracts and ultimately scale distribution.

Also briefly explored selling as an employer benefit, but didn’t feel confident on utilization or durability of contracts on selling to employers whose core competencies aren’t benefits.

Challenges & learnings

CI/CD

Main priorities on this project were product velocity and iteration. As a result, figuring out a good balanced system for CI/CD was a must, along with rigorous regression testing. Due to our limited user base, every shipped feature needed confidence it wouldn’t break in prod, since turnaround could be >48 hours with app store review.

On top of my standard system, I designed a playwright-based agent system to smoke test new features, click around the app, and screenshot screens in the on-device iOS simulator to iterate and fix glaring issues before manual review. Ended up catching many bugs, auth errors, and visual rendering issues which was extremely helpful as a small team shipping fast.

Onboarding

One of the earliest mistakes I made. Came off of building b2b products and saw onboarding as a “necessary evil” meant to be as simplified and minimal as possible. However, quickly realized that in a consumer world, onboarding is a make or break moment where you have the most new user eyeballs demanding you pitch them your app’s value.

Our product was inherently something that switched the systems that people were used to. Although we tried to keep UI/UX familiar, there was an inherent level of time and usage investment required. This actually required a more hyperpersonalized onboarding approach popularized by the new age of consumer apps like CalAI, Tiktok, Duolingo, etc.

In v2, I refactored our simple 3 screen flow to 50+ screens, and added in a mascot, animations, haptics, proof points, and EHR integration. However, in consumer health there’s an especially high bar for reputability and demonstrated value since questionnaires often asking for sensitive info. In reality at our product stage the ideal onboarding likely should’ve been somewhere in between.

UI/UX challenges

Our users were often working a full-time job and had kids on top of caregiving. It was really important to curate a very lean experience while still promoting in-app habits. Couple examples of where this manifested:

1. Checklists. Many of our users used printed or excel checklists to keep organized. My first intuition was that digitizing them could provide a reason to open and come back, but recurring feedback told us it actually felt like more work to adopt a new system.

2. Appointment Insights. Users were already overwhelmed with discharge instructions & lengthy medical documentation. Took a significant amount of tuning and evaling to generate recs / interpretation for a caregiver who is primarily concerned with what they need to do and why.

Marketing

Experimented across the board with paid ads, organic socials, SEO, cold-calling, partnership-led. The latter two had high hit rates, but timelines were slow and lots of drop-off to actually reaching end-users. On the other hand, our socials had good engagement which was encouraging, but rarely led to downloads. We were in the early stages of building the product, framed it as an “all in one” solution, and weren’t able to consolidate to a compelling core feature statement.

Increasingly, what we realized was “caregiver” was an umbrella term that provided us ease of use but families themselves did not ascribe to, making this a very interesting marketing problem to tackle. There were a lot of promising adjacent spaces like memory, longevity, etc. which were highly correlated and had a lot more consumer spend, while there weren’t really examples of successful products in the “caregiving” niche.