Voice-To-Text Optimization

Group Ride Function Iteration | UX/UI Design

About Group Ride

"Group Ride" is a significant iteration goal shared by Bryton's bike computer products. We aimed to increase the stickiness of long-term users and attract a wider range of users to expand the market.

Through integration with the Active App, this feature enables real-time access to ride information for cycling groups, helping riders obtain member locations, ride data, and messaging directly through the bike computer device.

Background

After the Group Ride feature came out to the market with the launch of Rider S800, it had limited device support and a small user base. During this period, we kept observing user data and conducting usability tests to optimize the voice messaging experience.

Usability Test Insights

Initially, we conducted several usability tests on the indoor training facility at office, involving colleagues with different ages and riding abilities from the staff cycling team. We used these observations to establish a usable operational workflow. However, during the subsequent on-road tests, we discovered that the workflow was totally not workable as expected.

Through 3 times on-road cycling tests and post-ride interviews with different users, we realized:

Even when the Group Ride feature was active during a ride, most users still kept their customized data pages on as the main meter screen.

The success of voice recognition greatly depended on the first three seconds of audio, and a convoluted workflow affected the user experience and, consequently, the success of voice recognition.

The hand controlling the bike might not be the hand users were accustomed to using for operating the bike computer device. This was especially challenging for them to operate the device with the non-dominant hand.

"Although it's convenient when it works, but practically it often fails, and the process of repeated operations to correct it makes me uncomfortable since it’s dangerous."

Problems

The initial design of the voice-to-text messaging feature had a complex process that was similar to the voice location search function used in navigation flow:

Member List > Chatroom > Voice Message > Start Recording > Speak Message > End Recording > (Wait for Recognition) > View Recognition Result > Confirm Sending

The original workflow required eight screens and at least five interactions (shapes in blue) to complete, which included two critical factors (shapes in green) that could significantly influence the user experience:
‍
■ The success of voice recording and recognition.
■ The accuracy of the recognition results.

These two elements had the most substantial impact on success rates of voice recognition and user’s feeling.
‍

Problem Definition

The Group Ride feature faces a significant challenge regarding hardware's audio recording capabilities. The first three seconds after starting recording is crucial for voice recognition success, but various factors like wind noise, ambient noise, and road traffic affect the accuracy and reliability of recognition results.

The original design was cumbersome during actual cycling. For most users, it was unsafe to focus on such a complex operation while riding and simultaneously stay focus on the road. Additionally, it increased the chances of recognition errors within the first three seconds, as users struggled to maintain a stable riding posture while interacting with the bike computer device.

Outcomes

After several usability tests and user feedback, we made the following improvements:

Member List > Chatroom > Voice Message > (Starts recording after a 3-second countdown) Speak Message > End Recording > (Wait for Recognition) > Automatically send the recognition result after an 8-second countdown.

These two adjustments reduced the workflow from at least 5 to just 3 steps(grids in blue) and improved the overall user experience and the success rate of the action. The changes were as follows:

A 3-second delay after clicking "Voice Message" to begin voice recording.
We implemented a 3-second countdown after clicking "Voice Message" to begin voice recording. This short delay allowed users to reposition themselves after operating the bike computer device with single hand, improving safety and reducing recognition errors in the first three seconds.
Auto sending out the recognition result after an 8-second countdown.
We defaulted to automatically sending the recognition result after an 8-second countdown. Users could tap the screen to send the message immediately or had 8 seconds to decide whether to cancel the message.
This improvement enhanced the overall messaging experience, enabling users to focus on their ride and road safety after confirming the accuracy of the message.
Differentiate the interface design between device states
This improvement enhanced the overall messaging experience, enabling users to focus on their ride and road safety after confirming the accuracy of the message.

Results and Takeaways

The workflow optimization in this phase includes simplifying the process and enhancing visual clarity, significantly reduced recognition failure rates and increased user willingness to use the function repeatedly.

+32%

Voice message usage

+26%

Voice recognition success

18%

Completed transmissions

+28%

Repeated use

Voice-To-Text Optimization

Problems

Problem Definition

Outcomes

Results and Takeaways

Thanks for Watching!

More Projects