Project in UX Design Sudio
Rate Your Lyft
Redesign Lyft's current five-star rating system
Lyft's current rating system tacitly encourages users to rank their driver using a five-star scale. When a Likert-scale system like this is used to give raises and even put someone's employment in question, precision is essential.
We argued that it's impossible to accurately measured feelings using numbers. In this project, we utilized different user research methods to explore a replacement for the current five-star rating system that Lyft uses to track and reward drivers in their system.
4 graduate students at Purdue
To understand the current riding experience and wear users’ shoes, we collected data from secondary research and interviews. Here are some interesting insights we got:
Five-star rating systems are inherently flawed because of the individual's perception of what a star is worth.
The number can easily be skewed by extreme scores
Riders only leave reviews/ratings when they feel very happy or very upset
Riders were biased by their past experiences and other people's ratings
Due to time constraints, riders sometimes rate the same score no matter how good or bad the ride is.
Then, we created an affinity diagram to analyze these emotions and actions.
One design opportunity we found is that Lyft's riders could only rate when the drive is over. It might halt the users from reaching their destination and lead to hasty reviews.
From here, we wanted to make sure that rating is part of the riding experience itself.
To keep our users' needs and feel in mind, we created two personas and scenarios based on our research outcome. Both of these Personas were under time constraints and were using Lyft out of convenience.
We ideated and sketched separately, then put our ideas together to find the best solution. Four design decisions were made after discussion.
1. Use 3 smiley faces instead of numbers
We wanted to solve the problem that data can easily be skewed by different people's scoring standards so we decided not to let people give numerical scores. Also, since people only write comments in extreme cases, we believe that the excessive band in the 5 smiley faces is not necessary.
2. Simplify the evaluation
The previous evaluation descriptions Lyft uses are navigation, friendliness, safety, and cleanliness. However, we thought both navigation and safety belong to driveability. From our interview, we also found users care more about safety so we use ‘safe’ to evaluate driveability.
3. Use thumbs up thumbs down to evaluate the three classification
We believed that the thumbs up/ thumbs down mechanism humanizes the ratings and can help people make decisions more intuitively and quickly.
4. Invite users to rate when the drive has 5 minutes left
We decided to send the user a notification/invitation to rate 5 minutes before the end of the drive. With more sufficient time and a relaxed mood, users might give much more effective feedback about the drive.
Design & Test:
Before creating a high-fidelity prototype, we made a quick paper prototype to gain some instant feedback.
Based on the result, we decided to simplify the rating page and stressed the signifier on the page.
Get notification 5 minutes away from the destination
If users don't want to rate at that time, they can click "Later" and go to the "Map" page
View the status of the car or Click on the "Rate" button to redirect to the rating page
After choosing an emoji, users can use 3 classifications to rate detailed aspects of the ride
On this page, users can see their rating and make a comment
6. Editing rating
Users can revise their previous rating by clicking the "Edit" button on the "Payment" page
We tried our best to redesign and tried to view the problem from a different perspective. We changed our design several times as we kept finding new problems and needed to do more researches. Yet, with more considerations, our designs were more complicated, we had to keep it simple. It was important to keep in mind that the chaos problem which seemed no answer to us will actually be done by users in 1 minute or even 30 seconds. The process was complex and iterative but I learned a lot from it :)