Product Designer, Contract
6 min read
Impakt
Impakt
How I designed Groups and a rating flow that helped an AI fitness app grow.
How I designed Groups and a rating flow that helped an AI fitness app grow.

5 months
Timeline, Nov 2024 to Mar 2025
Product Designer
Role, contract
CPO, CEO, CTO
Reported to
AI fitness app
Product
iOS and Android
Platforms
Figma, TestFlight
Tools
5 months
Timeline, Nov 2024 to Mar 2025
Product Designer
Role, contract
CPO, CEO, CTO
Reported to
AI fitness app
Product
iOS and Android
Platforms
Figma, TestFlight
Tools
Introduction
Introduction
Impakt is an AI fitness app that uses the phone camera to follow your movements and coach you through workouts. I joined a company of about 30 people as a contract product designer and owned features from the first idea to the App Store.
Impakt is an AI fitness app that uses the phone camera to follow your movements and coach you through workouts. I joined a company of about 30 people as a contract product designer and owned features from the first idea to the App Store.
My briefs usually came from the CPO in a sentence or two, such as “let users create fitness groups and invite each other.” From there I researched how other apps solved the problem, decided which mechanics fit Impakt, and designed the feature from low fidelity to developer-ready screens in the existing design system. During the build I answered the developers’ questions, and I was the first to test each release in TestFlight before giving the final approval to ship.
My briefs usually came from the CPO in a sentence or two, such as “let users create fitness groups and invite each other.” From there I researched how other apps solved the problem, decided which mechanics fit Impakt, and designed the feature from low fidelity to developer-ready screens in the existing design system. During the build I answered the developers’ questions, and I was the first to test each release in TestFlight before giving the final approval to ship.
This page covers the two features with the clearest outcome: Groups, and the rating flow behind Impakt’s App Store page.
This page covers the two features with the clearest outcome: Groups, and the rating flow behind Impakt’s App Store page.
At a glance
At a glance
177
177
groups created by users
25 → 233
25 → 233
App Store ratings worldwide, before the rating flow and today
4.7
4.7
average App Store rating today
8
8
localized App Store screenshot sets, used in more than 10 countries
App Store figures as of November 2026.
Part 1: Groups
Part 1: Groups
The brief was short: let users create fitness groups and invite each other, a bit like a WhatsApp group built around workouts. The bet behind it was gamification. If friends compete to get healthier, they get healthier and open the app more often. And since every group needs members, every new group was also an invitation to new users.
The brief was short: let users create fitness groups and invite each other, a bit like a WhatsApp group built around workouts. The bet behind it was gamification. If friends compete to get healthier, they get healthier and open the app more often. And since every group needs members, every new group was also an invitation to new users.
Research
Research
Before drawing anything, I studied Strava’s clubs, the closest match to our use case. Strava supports group photos, descriptions, public groups and challenges created by its own team, but no challenges inside a group. Competition between members became the space for Impakt.
Before drawing anything, I studied Strava’s clubs, the closest match to our use case. Strava supports group photos, descriptions, public groups and challenges created by its own team, but no challenges inside a group. Competition between members became the space for Impakt.

Three versions
Three versions
1
v1: a shared goal
Groups were built around a shared calorie goal per cycle. The group leader set how much each member should burn, members could nudge or cheer each other with one tap (“Don’t let the couch win today!”), and the member list was ranked by goal progress, then level, then join date. I also designed every management state: invites, broken invite links, removing members, leaving and deleting a group.
1
v1: a shared goal
Groups were built around a shared calorie goal per cycle. The group leader set how much each member should burn, members could nudge or cheer each other with one tap (“Don’t let the couch win today!”), and the member list was ranked by goal progress, then level, then join date. I also designed every management state: invites, broken invite links, removing members, leaving and deleting a group.

2
v2: simpler and more competitive
I reviewed v1 with the team, and the feedback pushed competition to the center. v2 removed goals and cycles, so creating a group takes a name and a description. The shared goal gave way to “Heroes of the week,” a ranking of the members who burned the most calories that week, and each group got four tabs: posts, leaderboard, chat and members.
2
v2: simpler and more competitive
I reviewed v1 with the team, and the feedback pushed competition to the center. v2 removed goals and cycles, so creating a group takes a name and a description. The shared goal gave way to “Heroes of the week,” a ranking of the members who burned the most calories that week, and each group got four tabs: posts, leaderboard, chat and members.

3
v3: two kinds of groups
The next round of feedback showed that the feature had two audiences with different needs. v3 splits them at the moment of creation: a Community, for coaches, influencers and classes who want to grow a following, and a Sweat Squad, for 2 to 6 friends or family members who hold each other accountable. I designed two directions for the creation flow so the team could compare them side by side. v3 shipped with Direction A.
3
v3: two kinds of groups
The next round of feedback showed that the feature had two audiences with different needs. v3 splits them at the moment of creation: a Community, for coaches, influencers and classes who want to grow a following, and a Sweat Squad, for 2 to 6 friends or family members who hold each other accountable. I designed two directions for the creation flow so the team could compare them side by side. v3 shipped with Direction A.

Outcome
Outcome
Users created 177 groups. Groups were meant to work as an invite loop, where two or three friends start a group and bring in others, and new users did arrive through invites. We could not separate invited installs from organic ones, so I can’t put a number on that part.
Users created 177 groups. Groups were meant to work as an invite loop, where two or three friends start a group and bring in others, and new users did arrive through invites. We could not separate invited installs from organic ones, so I can’t put a number on that part.
Part 2: Ratings and the App Store page
Part 2: Ratings and the App Store page
The problem
The problem
Impakt was live on the App Store but barely moved in the rankings. When I looked into it, the app had about 25 ratings worldwide and almost no new ones were coming in. Ratings matter for ranking and are among the first things a visitor sees, so I took the problem to the leadership team with a proposal: ask users at the right moment, and fix the store page they land on.
Impakt was live on the App Store but barely moved in the rankings. When I looked into it, the app had about 25 ratings worldwide and almost no new ones were coming in. Ratings matter for ranking and are among the first things a visitor sees, so I took the problem to the leadership team with a proposal: ask users at the right moment, and fix the store page they land on.
Three constraints
Three constraints
1
Apple shows its native rating prompt at most three times a year per user, so every showing counts.
1
Apple shows its native rating prompt at most three times a year per user, so every showing counts.
2
Google does not allow any question before its review prompt.
2
Google does not allow any question before its review prompt.
3
A prompt shown to an unhappy user invites a one-star review.
3
A prompt shown to an unhappy user invites a one-star review.
Timing
Timing
Timing mattered as much as the flow. The first proposal was to ask right after the third daily workout. The final rule asks users who completed three workouts in the past seven days, and skips anyone whose last workout was a beta workout. After a yes, the flow pauses for 120 days; after a no, for 45 days, down from a first proposal of 180 and 60.
Timing mattered as much as the flow. The first proposal was to ask right after the third daily workout. The final rule asks users who completed three workouts in the past seven days, and skips anyone whose last workout was a beta workout. After a yes, the flow pauses for 120 days; after a no, for 45 days, down from a first proposal of 180 and 60.
The flow
The flow
On iOS, after a workout, the app asks one light question: “Enjoying the app?” A yes leads to Apple’s native rating prompt. A no opens a short message form for the support team, so an unhappy user gets help instead of having to leave a public review, and we learn what went wrong. On Android, the native review card appears directly, without the question, at most once every 120 days.
On iOS, after a workout, the app asks one light question: “Enjoying the app?” A yes leads to Apple’s native rating prompt. A no opens a short message form for the support team, so an unhappy user gets help instead of having to leave a public review, and we learn what went wrong. On Android, the native review card appears directly, without the question, at most once every 120 days.

The store page
The store page
In parallel, I rebuilt the App Store screenshots and localized them into 8 sets for markets I chose from my own research. They cover more than 10 countries; the Spanish set, for example, served both Spain and Mexico. The team later translated the app itself into the same languages, and Spain became a new source of users.
In parallel, I rebuilt the App Store screenshots and localized them into 8 sets for markets I chose from my own research. They cover more than 10 countries; the Spanish set, for example, served both Spain and Mexico. The team later translated the app itself into the same languages, and Spain became a new source of users.


Outcome
Outcome
Before the flow went live, Impakt had about 25 ratings worldwide. Today it has 233, with an average of 4.7. I can’t separate how much came from the flow, the localized pages or changes after my contract ended, but more than 200 ratings have been added since.
Before the flow went live, Impakt had about 25 ratings worldwide. Today it has 233, with an average of 4.7. I can’t separate how much came from the flow, the localized pages or changes after my contract ended, but more than 200 ratings have been added since.
More from Impakt
More from Impakt

Feed interactions
A tap map for the social feed, covering what every area of a post does, plus report and unfollow with undo.

Streak repair
Users can spend earned coins to repair a broken streak. The price rises the longer they wait: 500, 1,500, then 2,500 coins within 72 hours.

AI instructors
Four instructor personalities with their own voice and lines, and rules for when they speak.

Quests
Campaign quests with tasks, points, a leaderboard and a tie-break rule.
Wrapping it up
Wrapping it up
What I learned
What I learned
A two-sentence brief is not a small task. Most of my work at Impakt happened between the CPO’s idea and the developer handoff: researching, choosing mechanics and testing them with the team until they held up. Groups changed shape twice because of feedback, and it shipped better for it.
A two-sentence brief is not a small task. Most of my work at Impakt happened between the CPO’s idea and the developer handoff: researching, choosing mechanics and testing them with the team until they held up. Groups changed shape twice because of feedback, and it shipped better for it.
What I would do differently
What I would do differently
Agree on measurement before the first screen. Both features were growth bets and both moved the numbers, but I can only show part of the evidence. Next time I would set up invite and prompt tracking from day one.
Agree on measurement before the first screen. Both features were growth bets and both moved the numbers, but I can only show part of the evidence. Next time I would set up invite and prompt tracking from day one.

