We hired a full-time anthropologist before our second engineer

Three months after launch, Verde had 12,000 downloads, a 1.4% retention rate at week eight, and an App Store review section full of quietly devastating feedback. "It feels like an app a man designed." "I cried when it asked me about my cycle in this clinical way." "Deleting."

We were building a women's-health app. The team was technically excellent. The product worked exactly as designed. The design itself was the problem, and not in a way any A/B test was going to fix.

I'd spent the first ninety days obsessing over engagement metrics that were telling me nothing about the actual relationship users had with the product. The metrics said churn. The reviews said hurt.

Verde started because I'd spent six years inside a digital-health unicorn watching dashboards built for clinicians, not for the people whose bodies were on the dashboards. I left to build something that started from the user's lived experience and worked backwards.

I raised a small seed off the strength of my pitch deck and my prior exits. I hired one excellent engineer. We shipped v1 in four months. The product did what I'd promised investors: cycle tracking, fertility windows, hormone-aware recommendations.

It also did what I had not noticed: it asked deeply personal questions in clinical language. It used graphs that made bodies feel like spreadsheets. It celebrated "regular cycles" in a way that made users with irregular cycles feel like errors.

The metrics told me retention was bad. The reviews told me retention was bad for a reason that no metric could see.

My investors were focused on engagement levers — onboarding flows, push notification cadence, A/B-tested copy. I tried all of it. Retention moved by a percentage point and then back down. The fundamental tone of the product wasn't getting better.

The pivotal moment was a conversation with a user who'd left a one-star review. I called her — she had left a phone number in her support ticket. She talked to me for forty minutes. She said, near the end: "Your app made me feel like a faulty machine."

I hung up and cried. Then I called my investors and told them I was going to make a hire that would be hard to justify on their next board update. I wasn't asking permission. I wanted them to hear it from me first.

I hired Dr. Yarima Soto, a medical anthropologist who had spent ten years studying how people talk about their bodies in private. She was our second hire after the founding engineer.

Yarima spent her first six weeks doing nothing visible to the company. She conducted twenty long-form interviews. She read every App Store review across our category. She built a vocabulary database — language patterns that users used spontaneously about their bodies, separated from the clinical language the apps imposed on them.

Then she sat with our designer and rewrote every screen of the app.

The new version did not show graphs by default. It showed sentences. "Your body is doing something different this week" instead of "Anomaly detected." "We don't know why and that's fine" instead of "Insufficient data." The cycle screen became a paragraph that read like a friend talking. The numbers were still there, one tap deeper, for users who wanted them.

We shipped v2 four months after Yarima joined. Retention at week eight moved from 1.4% to 11%. App Store reviews — the same place that had crushed me — became the strongest growth channel we ever had. By the end of the year we had 380,000 active users and a CAC under $4. Our biggest competitor copied the visual language within two quarters. They never copied the underlying research.

1. 1
   The metric isn't the problem; the metric is the symptom.

   Retention told us users were leaving. Only qualitative work told us why. A/B tests on the wrong layer of the problem are theater.

2. 2
   Hire the function the product needs, not the org chart investors expect.

   "Anthropologist" wasn't on any sample seed-stage org chart. It was the highest-leverage hire we ever made.

3. 3
   Language is product.

   In a consumer app touching anything personal, the words on the screen are not copy — they are the product. Get them wrong and nothing else matters.

4. 4
   One forty-minute call with a hurt user is worth a thousand survey responses.

   The user who left a one-star review and a phone number was the most generous teacher I've ever had.

5. 5
   The thing that's hard to justify on a board update is often the thing that justifies the next board update.

   Build the hire. Defend it later.