Voice User Interface

Alexa Skill

MY Role

User research, prototyping, VUI design

Timeline & tools

4 Weeks
VoiceFlow, Alexa Developer, Figma

the challenge

Creating a VUI for users to record sleep events during the night and eliminating the need to look at a mobile device.

Project Description

The goal of the project was to validate a hypothesis and create an MVP solution in a four week sprint. My hypothesis: 'people with insomnia need a way to track their sleep patterns through voice command' and I confirmed this through industry research and user interviews.

Sleep specialists ask patients for sleep information such as bedtime and nighttime awakenings, but don't accept information collected by a wearable as accurate. Patients don't want to manually record their sleep events as they avoid turning on lights during the night, which might wake them up more. I concluded there was a market fit for a voice activated sleep tracking app.

Outcomes

Alexa Skill

Intuitive and natural conversation paths for users to track their sleep events.

Data Tracking

Collected data gives users the power to make informed sleep decisions around their habits to improve sleep quality.

Research

To begin the project I wanted to find out why people with insomnia frequently track their sleep patterns. Through user interviews I learned about tracking methods, discovered pain points and developed an understanding of how people relate to their sleeplessness. I found people feel a lot of stress when they are constantly thinking about how to improve their sleep quality.

User Interviews: Key Findings on Sleep Tracking Habits

01

Good data is key to making good health decisions

"I would make almost any lifestyle change to improve my sleep quality. But, if it’s a big change I want to make sure it has a chance of working. I don’t want to just be trying new habits blindly"

02

Tracking sleep during the night without looking at a screen is imperative

"At night I do everything I can to avoid light. I ask Alexa what time it is instead of looking at my phone, I wear an eye mask, and I make sure to never turn on the light."

03

Sleep tracking can be stressful, because not sleeping is stressful

"Laying in bed at night, awake, might be the most stressful thing I’ve ever experienced."

Looking at the competition

I did a competitor analysis of the top five sleep tracking apps and found tracking using a wearable is common, but voice-activated sleep tracking is not. Most sleep doctors only accept self-reported sleep data from their patients as they view information collected by a wearable as inaccurate. As people with insomnia avoid light exposure during the night, manually recording bedtime and nightime awakenings on a mobile device isn’t an option.

Define

Outcomes from user interviews and market research showed Amazon Alexa is one of the most commonly used smart speakers in the bedroom. As such, I decided creating an Alexa Skill was the best format to address the project problem.

Problem Statement

"I need a reliable way to track my sleep during the night without looking at a screen."

Personas

Just as two people talking to each other have their own personalities, so do the user and the system. Creating a robust system persona helped me empathize the user-system interaction. Market research shows half of smart speaker owners report speaking to their device feels like talking to a friend or another person. A strong system personality makes this happen; just look at Luke Skywalker and R2D2.

Information Architecture

Designing for VUI best practices meant creating conversational paths similar to how people talk in real life. Teaching Hypnos how to communicate as humans do involved defining all possible conversational paths. Using a list of sleep events users want to record, which I compiled during user interviews, I created a chart of eight possible user requests and the conversational paths they trigger.

"The value of CTAs and button states in GUI is just as important as well-defined intents in VUI"

Building for Alexa

To define information architecture and conversational paths, I followed AWS language structure
Skills are like apps for Alexa.
An intent is the action that completes a user’s spoken request.
Utterances are a list of spoken phrases that activate an intent. When going to bed a user might say, “It’s bedtime,” “I’m going to bed,” or “I’m getting in bed.”
A slot is variable information within an utterance. Slots help Alexa have more natural and complete conversational flows.

User Flows

The goal of the sprint was to create an MVP but the larger goal of the project was continued product development. This, it was important to create deliverables with enough information for the entire product team, especially the project engineer. My first user flows were specific to the product I would make in Voiceflow. For a smooth hand off to the engineering team I created additional in-depth user flows for intents showing the actions of the user, and system, and the relationship between them.

Design

To accomodate different conversational paths, both single- and multi-turn, I developed Hypnos as a group of intents which can be navigated to directly using utterances or indirectly through the main Hypnos welcome message. Using an open format, instead of fixed paths, allows the app to be easily navigated by new and existing users alike. Each intent records a specific sleep event such as bedtime, nighttime awakening, and recording sleep quality.

Building conversations accessible to new users and adaptable to returning ones

Prototype & Testing

After designing and creating a prototype I needed to test if the product would work for a real user. I chose moderated in-person testing in order to observe how users interacted with their smart speakers. The main usability testing objectives were designed to find gaps in conversation, broken paths, and undefined utterances and slots.

Objectives

  • Evaluate how users record sleep events and respond to prompts
  • Assess users ability to navigate conversational paths

Methods

In-person moderated testing

Tasks

  1. Log your bedtime in Hypnos
  2. Log your awake for the day and track your sleep quality as a four (out of five).
  3. Learn about what sleep events can be logged in Hypnos.

Affinity map and priority changes

I was surprised to find how many new utterances users tried for intents during the usability testing. There are many ways to say you’re going to bed and how do you differentiate between a nightime awakening and recording you’re awake for the day? Implementing changes to utterances will increase ease of use of the interface.

Reflection & Next Steps

VUI is an opportunity space. Feedback from usability testing was positive, and it reveals users comfort levels engaging with smart speakers. I was surprised to find most users engage with Alexa for primarily single-turn interactions (ie. setting a timer or playing a song) and are less familiar, and comfortable, engaging in a back and forth dialogue. Given this discovery, an onboarding feature for new users to the app would be valuable to familiarize them with the conversational format. Further, it indicates an opportunity for growth within the market.

What I learned

VUI is a huge opportunity space

Users enjoy interacting with voice assistants but have a very limited understanding of what they’re capable of.

Focusing on currently available tech for the MVP

Utilizing new technologies and understanding what might be available in the future can be the source of many ideas. Staying grounded in the real world and making an MVP with currently available technologies is essential to product development.

Prioritization is key

I frequently referred to project goals to keep on track and evaluate the viability of adding more features. Important features that didn’t make the MVP were added to a v2 list.

Next Steps

  • More usability testing & iteration
  • Implement logic in conversational paths to decrease the users cognitive load
  • Develop a multi-modal plan & UI to display data recorded in Hypnos
  • Create a way for users to share tracked sleep data, specifically with their doctors

PS

I discovered a few humorous perspectives on Alexa during my research that gave me inspiration. A view on what interactions with Alexa could look like in the future, with Alexa Silver, and an example of how AI natural language processing has a lot of space to improve accuracy in comprehension with a scene in an elevator.

Other projects

UI Design | UX | User Research

End to End Mobile App
UI Design | UX | User Research

Website Design
Voice User Interface Design

Alexa Skill

Contact

@2021 Claire Waugh