Voice User Interface • Daily Haiku
Amazon Alexa Skill
Daily Haiku
Voice Design Case Study
My Role: UX Designer and Voice Designer
Solo project
New York 2019
Objective & Challenge
Create an Alexa skill that reads haiku upon request. Users can request various kinds of haiku poetry, such as romantic, funny, modern, or traditional, and request to repeat the previously read ones.
The challenge of the project was to create an amusing and easy-to-navigate voice experience that offers a simple haiku reader upon request.
The 5 W's of Daily Haiku
What [does it do]?
A database with a variety of randomly read haiku poems upon request. The collected haiku poems are available on public websites that will be listed below. I do not own any of them. The skill is created to promote the beauty and wisdom of haiku poetry, entertain, and share light humor.
Who [is this for]?
Anyone who is interested in hearing haiku poems. The average age of the target audience for this project cannot be determined as there are no age limits, but the overall users belong to these three categories:
Daily users – the users who love to listen to haiku
Occasional users – the ones that, from time to time, like to hear a nice haiku, or two
One-time-only users: the ones that want to listen to a haiku sample or simply test the voice skill
When [can it be used]?
The skill grants unlimited access during the day, so users can have their wise words for a bright morning, enthusiastic poetry during the day and also calming words before bedtime. The user should ask Alexa to launch Daily Haikus, and the world of haiku becomes available to the users
Where [can it be used]?
Anywhere users have access to the Alexa app.
Why [would someone use it]?
This skill covers various haiku topics, including profound, funny, traditional, and modern, and inspires users to travel deep into their thoughts and empower the creative energy flow. It is also playful and fun.
Proto-Persona & System Persona
As previously mentioned, the target audience for this project was not specifically defined, as the target group can be anyone who likes and wants to hear haiku poems, so I decided to create an initial proto-persona named Ayuko.
On the other side, creating a System persona and its character was equally important, even though it has a few character lines: understand the user prompt, find the requested poem type, read a specific type of haiku poem, and repeat it upon request. Lumi is the basic system persona I developed in order to navigate users though the Daily Haiku experience.
Needs & Goals
Frustrations
Ayuko wants to learn about her origins and culture
Wants to connect English and Japanese literature and explore the connections between the two
Always tries to find deeper meaning in ordinary things
She is fascinated by the two very different cultures she is equal part of
Ayuko knows that she could learn more about her culture if she could visit Japan, but currently can't afford to travel there
She works on project of combining English and Japanese literature and mutual influence
Her current Japanese is limited, so she wants to explore haiku in English, and understand the meaning
Everyday Activities
Reads books, magazines, and everything she finds interesting in order to improve her Japanese vocabulary
When taking a study breaks, she likes to do something useful but relaxing, instead of just scrolling though her phone
Loves educational, empowering podcasts, especially the ones about travels and exploring new cultures
Polite • Engaging • Informative • Kind • Warm-hearted • UNDERSTANDING • HELPFUL • haiku poetry lover
Lumi is the Daily Haiku system person and the haiku connoisseur. As traditional haiku poetry comes from Japan, I decided to combine it with a modern-age character, as the user interests can be quite broad, and the communication is via voice assistant, the device of the modern age.
The important part of developing the system persona was introducing emotional intelligence in the system’s character by understanding the user’s needs. For example, if a user wants to hear a funny haiku poem and then asks for another one, Lumi will understand the user wants to hear more funny haiku poems.
Research & MVP Validation
The research consisted of four parts:
The first part of the research was about understanding the opening utterances on how to wake up and interact with the voice skill. To understand how different voice assistants work, I ran a competitor analysis of different voice skills and mobile apps that use voice/ sound for interaction (Bixby, Siri, Cortana, Google Assistant, ATT app, App Lock (Face/Voice recognition), Google translate).
The second part was related to prompt variations. I combined the competitor analysis and heuristic evaluation to understand the prompt variations. I analyzed Alexa's skills: Common Knowledge, TED Talks, Eight Hours of Sleep, and Make Me Smart
The third part of the research followed after creating a database of multi-line responses read with human-like pauses and breaks, custom help messages that help users recover from errors, and relevant utterances for the main intent and content access. The third part of the research consisted of remote usability testing with 5 users and asking them to complete six simple tasks:
wake up the Daily Haiky Skill,
request a haiku poem
request a new type of haiku poem
request to repeat a haiku poem
ask Daily Haiku to read the news instead of haiku,
request to return to the main menu
The testers’ feedback was crucial for improving the skill utterances and providing appropriate responses. The following pain points were discovered:
The skill did not recognize new users if the skill is already used on one Alexa device, as there is no way of creating a specific user profile, so it would sometimes read the same haiku poem a few times.
Test participants found it frustrating to say “Next” or “Another one” every time they wanted to hear a new poem, and they wanted skill to understand when they say “[tell me/read] More” or “[read a] New one”
Test participants couldn’t have more interaction with the system and chose “random” as a type of haiku poetry.
Finally, the fourth part of the research consisted of another round of usability testing with 5 new testers and measuring success rate using the level of success on the above-mentioned tasks.
All five participants agreed that the skill, even though very simple, is useful because it delivers the expected – the haiku poems, without additional requests from the users, such as sign-up or subscription.
On the other hand, the list of possible utterances was enriched with some additional examples provided by test participants, and some of them provided suggestions for future development. After quick polishing, the Daily Haiku voice skill was submitted for certification
Current Analysis & Future Development
By tracking users' activity such as:
the unanswered prompts that lead to help messages,
the duration of users’ interaction with the skill,
and how frequently the skill is being used by existing users vs new users,
I learned the system persona learned to react to different triggers according to users' behavior. Different user behavior has taught me different user goals and helped me create new ways to engage with the skill, and I introduced additional approaches to the user’s requests.
Returning users are recognized, and according to their activity, the system offers a warmer opening message. The greeting is less formal, and one poem is being read before the previous confirmation or request from the user.
For returning users, the prompts are changing, and the users should hear different intro prompts every time they launch the skill.
After reading the poem, the system should always ask users if they would like to hear more haiku poems. Depending on how many times the user proposed the new poem, the system can read only one poem or three in a row before asking if a user would like to hear more.
For future development, as new poems are being added (and the current database contains around 250 poems), the poems should be analyzed and sorted into easily recognizable categories. The list will be prepared, and then using the Card Sorting technique, users can sort them and create appropriate categories.