Sogou VUX Improvement

During my UX design internship at Sogou in summer 2018, one of my projects was enhancing Sogou App’s voice experience. Sogou App is a search App that facilitates browser search, news feed, novel reading, and multiple app-based functions.

Role: UX designer intern

Practices: VUX design, content strategy

Project Background

As part of the VUI improvement project, I was in charge of adding personalities to the voice assistant's feedback, making the interactions intuitive and delightful.


  • Understand the Users
    • Research user journey and discover pain points based on contexts
  • Understand the Product
    • Determine voice assistant’s personalities based on Sogou’s marketing strategy
  • Understand the Industry
    • Follow VUX design principles established by industry leaders


I identified three scenarios for voice interaction, the corresponding feedback types, and to what extent personalities should be displayed:

Ideation & Prototype

With the feedback types sorted out, I started to brainstormed feedback prompts that are delightful and with personality with my team. We went broad and creative on making the feedback delightful.

For error handling, we focused on clearly notifying users what went wrong and highlighting direct CTA in GUI design. For example, to enable the microphone, we added a “go to setting” button in the chatbot.

For confirming that a task is completed, we aimed to be as concise as possible so as to reduce users’ distraction. As for UI design, we used toast for presenting the confirmation note.

For chatting, we decided to direct users to a “Chatting Mode”. As for chatting prompts, we would use a “chatting library” that Sogou has previously created. We also added a Wangzai AR to indicate users have entered the “chatting mode”.


We conducted a round of "Wizard of the Oz" testing with colleagues at Sogou, and discovered that:

  1. Details of technical failures confuse users. They care more about what they could do to handle errors.
  2. Users expect concise feedback, but they also expect more conversational dialogues like real-life communications.
  3. When encountering errors, users expect Wangzai to apologize because they view Wangzai as an assistant that provide service.

Accordingly, we improved the error handling feedback by grouping errors scenarios based on users’ experience instead of technical reasons. We also made sure not to blame the users by adding “sorry” in every error case.

In addition, to make the confirmation prompts more conversational, I added random selected transitional words such as “Okay, sure, and already”.

The chatting mode is eliminated from this version of design at the iteration stage due to resource restriction.

Final Solution

The newest version was launched in November!