Her alarm goes off at 5 AM and there is no snooze. Breakfast for the kids, get them to school, then out to client jobs: two in the afternoon, one in the evening, one still unconfirmed. If that fourth one lands it's another forty dollars. She will not be home when her children are.
That was the person we were building for at Honor, and I opened a piece about it that way in 2015 because I could not find an honest way to start with the software.
What I want to write about now is narrower and, I think, more useful: what it actually does to your design practice when the people you're building for are nothing like you.
Everyone was a caregiver
We had a rule: everybody at the company did volunteer caregiving shifts. Real shifts, not a workshop or a ride-along. Cleaning bathrooms and kitchens, organizing family photographs, heavy work in gardens.
The stated reason was empathy, which is the word everyone reaches for and which I've become suspicious of. Empathy as usually practiced in tech means imagining harder. You sit in a room with a persona on the wall and you imagine.
Doing the shifts wasn't imagining; it was collecting facts you would never have thought to ask about. That a phone lives in a pocket that is frequently wet. That the moment you need to record something is the moment both your hands are occupied. That the emotionally significant part of the visit is often the ten minutes of conversation, which no scheduling system has a field for.
Some of what we learned validated our assumptions. A useful amount contradicted them.
Empathy is not a feeling you generate before designing; it's a set of facts you don't have yet.
The landline
The decision I still cite most often is that engaging with Honor had to work as well over a landline as over a smartphone.
That is an unglamorous sentence and it was an expensive commitment. It constrains everything downstream: you cannot assume a screen, so the model of what's happening has to survive being spoken aloud, which forces a clarity that a UI will happily let you avoid.
And it came from a specific, correctable mistake. Our early assumption was that our customers would mostly be adult children arranging care for a parent. When we went to market we found roughly half the people signing up each day were arranging care for themselves or a spouse.
That's a different person with different hardware and a completely different emotional posture. The adult child is solving a logistics problem. The person arranging care for themselves is negotiating with a change in their own independence. Same product surface, entirely different thing happening.
Against designing down
The trap when you build for older people is to produce the I've-fallen-and-I-can't-get-up-ified version of an ordinary product. Bigger type, fewer features, primary colors, an implicit apology.
What we found was that people much preferred products designed with context and built on systems they already knew, over products that had been designed down to them. I don't think the failure is insufficient accessibility so much as the condescension that gets encoded when you treat a demographic as a deficit.
Which is where the three words came from, the frame we ended up using instead.
Joy. Most well-designed products serve their purpose with simplicity and obviousness, but the good ones also carry something emotional. For the Care Pros that mattered enormously: to feel their profession was regarded seriously, to have tools that adapted to their lives rather than the reverse. For families, joy is usually the emotion furthest from the situation, which is exactly why its absence from the software is so conspicuous.
Comfort. Not softness but the reduction of uncertainty. People arranging home care mostly don't know what to expect, and are frequently caught between an aging parent and their own children. Transparency is how you get it: visit summaries did more for people's peace of mind than any feature we shipped that looked more impressive in a review.
Grace. Roughly ten thousand people turn sixty-five every day in the United States (that was the figure when we were building, and it has not gotten smaller). We treated that as ten thousand daily reasons to change how we talk to our elders. Society lumps them into one infantilized category, and that category is us, a few years out. A group as varied in condition and taste and temperament as any other.
What transferred
I've since worked on rideshare, on writing software, on presentation tools, none of which involve anyone's grandmother. The thing that carried over is not the three words but the operating assumption underneath them.
Which is: your intuition is a local instrument. It's well-calibrated for people like you and gets less reliable the further you go from that, and it gives you no warning as it degrades. It feels exactly as confident either way. The only correction is contact with the actual situation, and the only reliable way to get that is to go and do the thing, badly, yourself.
Being able to build for people outside the Valley's homogeneity was a humbling experience, and I mean that in the narrow sense: it lowered my estimate of how much I could work out from my desk.
I've never got that back, and I don't want to.
Where the shifts stop working
I have described a practice that worked, so it is worth saying what it costs and how it fails, because I have seen both.
The cost is straightforward. A day of caregiving is a day not spent building, and at a company of any size the arithmetic gets hard to defend to whoever is watching burn. We could do it because we were small and because the founders had decided it mattered, which is another way of saying it survived on the same footing as any other unfunded conviction.
The failure mode is subtler and I think more common. The shift becomes a ritual. People do it once, early, tell the story at onboarding for the next three years, and stop going without anyone noticing. What is left is a company that believes it has contact with its customer because it once had contact with its customer, which is worse than knowing you do not, because it forecloses the question.
The tell is when the anecdote stops updating. If the story your team tells about your user is the same story it told two years ago, nobody has been out recently, whatever the calendar says.
What I would build now is duller and more durable: a small standing obligation rather than a founding legend. Some hours a quarter, for everyone who makes product decisions, logged the way any other commitment is logged, with the findings written somewhere the next person can read them. Less romantic than everyone doing shifts in year one, and it survives the year when nobody has time.
I was part of the founding executive team at Honor, leading product and brand from seed through Series C. Portions of this revisit "A year of Honor," published December 2015. The counter-argument worth taking seriously: immersion doesn't scale, and at some point you need research infrastructure rather than everyone doing shifts. True, but the shifts are what tell you which research to commission.
© 2026 Renato Valdés-Olmos