How we designed for Voice at Cuddle.ai
At Cuddle, we recently integrated voice based questioning which enables users to ask questions about their business without typing. I’ve completed a month at Cuddle and one of my first tasks was to design for voice. The need for voice search emerged during discussions with potential clients at conferences, sales demos and multiple internal discussions.
As per Gartner’s predictions, by 2020 about 30% of all searches will be done without a screen. Google Home and Alexa, the voice assistants were popular gifts for Christmas in 2017. The trend of voice is upcoming. However, designing for voice is still ambiguous. Playing around with voice interfaces might be a little frustrating for first time users. Here is a video clipping of a grandmother trying out Google Home, that was presented to her during Christmas by her grand children that went viral.
How can we design a better voice experience to bring pieces of information faster and make it feel seamless? What are the right visual and verbal cues to show the user while interacting with voice user interfaces? How can we handle errors in an elegant manner without a frustrating experience? These were our initial thoughts and apprehensions as we went about designing voice at Cuddle. The following below is how we designed for voice —
- Research
- User Scenario and Flow
- Design + Initial Prototype
- User Testing and Insights
- Reiterating designs after Usability Testing
1. Research
In the recent times, quite a few apps now help you search through voice. To begin our research, we observed how other app interfaces like Google Neighbourly, Google Assistant, Siri, Whatsapp and Ganna to understand the basic user flow for voice. We also looked at apps and interfaces for recording voice in general to understand what kind of cues are users familiar with.
Designing a voice search for enterprise has challenges of its own. The nature of questions asked varies from business to business. What kind of questions do we enable our voice search for? How do we onboard users to ask the right questions? These were a few concerns during the process. Based on the available resources and time, we designed a flow for a user to ask certain type of question about his business.
2. User Scenarios + User Flow
We identified and address three scenarios where users would prefer to search by voice rather text
Scenario 1 — The users hand’s are occupied to perform another activity.
Scenario 2 — Presenting your business stats to an audience.
Scenario 3— The user requires a quick update about a particular metric.
3. Design + Initial Prototype
After a few iterations, we prepared the flow and screens for development implementation.
Designing for handling errors : With voice, there can be instances where the system will not be able to interpret the words correctly. We needed to handle such cases without causing frustration for the user. We enabled editing the question with a tap, incase the system wasn't able to interpret the question as intended.
4. User Testing and Insights
It was essential for us to test our prototype outside our team. The goal of this exercise was to understand how users perceive voice on mobile and to test how interact with our prototype for voice.We conducted the test with 7 users within the age group of 22 to 30 years. The task was to ask a question through the voice search on the Cuddle app.
Reiterating Designs over UT
Based on our observations, we jotted down the possible actions for improving the interfaces for better communication.
Discovery of voice search: Participants took time to find the mic icon on the screen. Introducing the voice search below the search bar when the user begins to type would help in gaining better visibility. An on-boarding showing what all the participants can do through voice would reduce initial friction.
Sound Notifications : Providing feedback to the user via audio feedback
Introduce a button for editing the question: Question to edit was not clear. Introducing a button would make the action clear.
Notification for ‘tap when done’ action to be shown in 3 seconds: ‘Tap when done’ notification needed to appear sooner. We also figured that multiple nudges were required to show this communication above, as tapping on the stop button wasn’t intuitive.
Other Observations : Users preferred to address the app the way google does it by saying ‘OK google” and then asking a question. 3 out of 7 participants expected the system to respond immediately after asking a question. 2 out of 7 participants did not read the textual content on the screen. They focused on icons, colours and and graphics for further cues.
Prototype
Here is the final prototype post user testing.
Thank you for reading. If you found this helpful, please show some support with *claps emoji* and share this with others. If you’ve designed for voice, I’d love to hear about your experiences, challenges and wins.
A big shout out to Hardik Chandrahas for the all video during the UT sessions and to Amit Das for guiding me to write my first blog post! :)
