Published day-7 retention benchmarks sit at 10–13% across all apps, and 7–8.5% is considered good for health & fitness apps. Our live beta holds 8% at day 7. This post has every number behind that: what actually retained people (not what we guessed), the button we never expected to matter, and the plan for the real fight, day 30, where good fitness apps hold just 3.5–4%.
One Goal Sherpa is a goal-setting app we build and run ourselves, aimed at women between 40 and 60. It is live in beta with real ad spend, which means real strangers, real churn, and numbers we cannot argue with. This is the log of what the data said this round and what we are building because of it.
What is a good day-7 retention rate for a goal app?
Across all app categories, the published day-7 median is about 10–13%. For health & fitness apps, the closest published category to a goal app, industry benchmarks call 7–8.5% at day 7 good. After fixing onboarding, we hold 8%, at the top of that band. The honest news: out of every 100 people who sign up, about 92 are still gone within a week. Here is the curve.
In the earliest version, people performed one ritual and left. Fixing the onboarding is what got the curve to where it is now. The next fight is day 30, where the floor drops fast: good fitness apps hold about 3.5–4% at day 30, and plenty of categories fall to 1–2%.
The feature that was quietly retaining people
During onboarding we ship something called the Map: each day the app has you fill out one new thing, and it builds toward a visual map of your goal that you can interact with. Because of where it sits in onboarding, 91% of people complete it.
Our assumption was that the Map was the hero of the product. The data said something more specific: the Map does retain people, but two in three users never see it again after onboarding, because it lives where nobody looks. The feature was not failing. The placement was.
"The Map is the hero." Reality: the Map retains the people who can find it. Most people could not find it. Next sprint moves it somewhere unmissable.
The button nobody was supposed to press
After 13+ years of shipping products, one pattern keeps repeating: the features you think are cool are rarely the ones users love, and the ones you almost cut become the hits. Two examples from this round.
First: the most-tapped feeling button in the entire app is "Overwhelmed." Not motivated, not energized. Overwhelmed. That single data point now shapes the product. A goal app that adds work to an overwhelmed person's day loses; every flow has to reduce load, not add it.
Second: when you complete a goal, the app asks "How did that feel?" I thought it was a dead question and was ready to cut the box. Then we pulled the numbers: out of 203 completed goals, 64% of people wrote real reflections in it. And it gets better.
Today those reflections go into a black box. We read them; the app does nothing with them. That is the biggest unbuilt lever in the product, and it is next on the list.
What retains users in a goal app, by the numbers
- Sessions that finish in under five minutes. Get in, set the goal, get out. The daily return loop lives or dies on this.
- A visible artifact of progress. 91% of users build the Map when it is put in front of them.
- Reflection at the moment of completion. 71% versus 49% weekly return is the strongest split we found.
- Reminders on a channel people actually see. SMS users return 2.6× more, but you cannot force the opt-in, so email and push are coming next.
- 28% of new users set a single goal and stop. Features that require discovering go unused. If it matters, it has to be unmissable.
The plan for day-30 retention
Round three of the build starts now, aimed at one number: how many people are still here at day 30. Four moves, straight from the data above.
- Surface the Map. Two in three users never see the thing that retains them. It moves to the center of the app.
- Close the reflection loop. 64% of completers are already talking to us. The app starts talking back, so the conversation continues the next day.
- Keep the daily ritual fresh. A new format each day, so returning never feels like a rerun.
- Premium polish last, on purpose. We prove which behaviors matter before we make everything pretty. Current UI is fine; finished it is not.
The two rules this build runs on
Rule one: build for the behavior the data shows, not the behavior my gut says is there. Every surprise in this post is a place my gut was wrong.
Rule two: never trust one query. One of our retention numbers came back 20× off. Every number that touches a decision now gets checked two ways: one AI model pulls the data, a second, different model verifies it independently, and if they disagree they argue it out until they reach consensus. Only then does the number drive a build decision.
The whole thing runs on one loop: build, test, analyze, iterate. Round one was onboarding. Round two was the six-minute session. Round three is day-30 retention. It will not be perfect, and that is fine. The data decides, and we keep digging.
Frequently asked questions
What is a good day-7 retention rate for a goal-setting app?
Published benchmarks put the day-7 median at 10–13% across all apps; 7–8.5% is considered good for health & fitness, the closest category to a goal app. Our live beta holds 8% (37% day 1, 23% day 2). By day 30, good fitness apps hold 3.5–4%, and many categories fall to 1–2%.
Why do users quit goal apps in the first week?
In our data, early users performed one ritual and left. The most-tapped feeling in the app is "Overwhelmed," so anything that adds work loses. Sessions that finish in under five minutes retained best.
What features actually improve goal app retention?
Three things moved our numbers: a progress artifact users build during onboarding (91% complete it), a reflection prompt at the moment of completing a goal (71% vs 49% weekly return), and SMS reminders (2.6× more returns).
How do you verify product analytics before acting on them?
Never trust one query; one of ours was 20× off. We run every decision-grade number through two different AI models independently, and they have to agree before the number is used.
The point of all this
This is what an AI department looks like.
Data pulled and verified by AI agents, decisions driven by real behavior, and a build loop that never stops. We run our own products this way, and we install the same system inside client businesses.
Book a 20-minute fit call