Golf AI Swing Analysis Redesign
May - August 2020
In October of 2019, I joined Golf AI, a mobile app startup focused on using machine learning technology to help golfers improve their swing. The first project I worked on revolved around improving the UI of the swing analysis screens at the time. However, later on we realized that we should have taken a closer look at the swing improvement process as a whole. So, starting in May the UX Researcher (Darin Lee) and the Product Design team (Anmol Arora, Pooja Yadav, and I) worked on redesigning the swing analysis from the ground up.
- Survey Design
- Journey Mapping
- User Testing and Iterating
- Product Strategy
- Developer Handoff
Understanding Our Users
As the startup grew more data-focused, we noticed low amounts of total users and sales. We wanted to revisit our main feature, the instant golf swing analysis, to see if it was providing as much value to our target users as we thought it was. However, we realized that we didn't have a comprehensive idea of our target users. We created and distributed a survey, led by Darin, to find out who should be our target users and to learn more about them. We reached 89 golfers and 16 former golfers and found the following insights:
Golfers with worse swing scores are more likely to use the app
Golfers want help from experts to be personalized to their own specific golf swing.
Visualizations are Valuable
Golfers value being able to visualize their golf swing and what they did wrong
Golfers want to know that they can trust the app's recommendations; how does it work?
We realized we needed to shift our focus to beginner to intermediate golfers. By doing so, we had a better focus on our target users' needs.
Current App Evaluation
With our user group in mind, we were able to re-evaluate what we had designed before. In doing so, we noticed many areas of improvement:
There was too much information for users, especially beginner golfers, to process. The next step for users to take is not clear.
No Focus On Trust
There was not enough emphasis on building trust with the users. The information button is too small and too close to the view video button, and is too low priority in this page.
Not enough context
Golfers can see their swing compared to an ideal swing side by side, but there is no context given to tell the user what they should be looking for. The specific area is also not explained, which a beginner golfer might need.
There is only one tip given per area of swing, and the tips were not very personalized to the user. As a result, they are not very helpful in fixing a golfer's swing.
The Big Problem
After looking at the survey results and our current app, we made the following how might we statement:
"How might we inform beginner golfers about the most significant mistakes in their swing and effectively give them the resources to improve?"
After creating this how might we statement, we sketched our ideas out then started working on low fidelity prototypes. We then tested the prototypes over Zoom, iterated, and retested.
Low Fidelity Iteration 1
User Testing: First Round
In our first round of testing, we wanted to focus on the following things:
Is the function of the help button clear to the users?
How useful is the format of overview, tips, and drills?
Do users care about areas that need work apart from what needs the most work?
How important are best areas to users?
After remote usability testing with three users over Zoom, this is what we found:
Users wondered what the swing score meant and how it was calculated, and didn't immediately know that the question mark icon would help them understand. We needed a clearer way to give users information about the swing score.
Users liked seeing what area they were best in. "I like seeing my best area, so I'll try not to mess that up."
Users always went to action plan first before pressing "How do I improve this?" This changed the way we thought about the action plan, and made us realize the action plan should be a larger part of the user's flow.
Users thought the format for the pages of specific areas was intuitive and easy to follow. However, we noticed that users would tap on the toggle bar to switch pages, so we found that the bottom buttons were unnecessary.
Low Fidelity Iteration 2
User Testing: Second Round
We tested three more users using the same method, focusing on the same set of questions we had during the first user test. This is what we found:
The new button to learn more about the swing score was a lot more effective and intuitive to users. Users would wonder aloud what the swing score meant, then tap on the "How is Swing Score calcultaed?" button.
Users liked the idea of having an action plan to work on. However, we realized that we didn't address the problem of not being able to visualize their best area.
The video in the tips section changes depending on the tip selected, but users were not pressing on the different tips, which we did not notice in our first set of user tests.
We transitioned into creating high fidelity iterations after the second round of user testing. After going through a few different styles, this is the final prototype that we designed:
You can find the link to the interactive prototype here.
After implementation and release on the Apple App Store, our sales in the following month increased by 61%!
This project was very rewarding for me because I was redesigning the experience that I had previously worked on in October 2019. Through this project, I was able to see my growth as a designer. However, designs are never finished, and we already have ideas we are trying to work into our current flow. Feel free to ask about what's coming to Golf AI!