Alexa Skill Design

I have omitted confidential information in this case study. All information is my own and does not necessarily reflect the views of Arrive.

 The Problem

The ParkWhiz Alexa skill started as a side project for an engineering intern, with the goal of allowing users to find a parking facility. After a few engineering focused iterations, we quickly realized that the first iteration of the ParkWhiz skill was becoming over-complicated, containing long user flows and clunky content. The ParkWhiz Alexa skill was prioritized with the launch of the Alexa Auto device. I became involved in the  project in 2019, and set out to create the best in-vehicle experience possible. 

Challenges

Designing for Alexa came with its own challenges. I needed to work closely with  engineering to  discover what was and was not possible. Documentation was not as easy to understand as Apple’s  Human-Interaction Guidelines or the Material Design Guidelines. Additionally, at the time we started building the skill, there were no wide-spread UX best practices for creating user flows and for displaying functionality and all content available within the skill. I established my own patterns for visualizing user flows for voice. 

Getting started 

The team involved in the end-to-end development of our skill included a PM, a dedicated voice engineer, an Account Manager, and yours truly. We also worked closely with a representative from Amazon throughout our continued progress and iterations. Together, using Amazon’s voice best practices, we established the following design tenants and goals. 

Design Tenants

• Better than the Alternative - reduce friction to find, navigate to and purchase parking 

• Build/Maintain Customer Trust - provide the best option for the user

• Two Way Doors - easy to try but easy to back out if you find a better alternative

• Global Coverage but Graceful Failure - go wide and then go deep

• Minimize User Inputs - use what you already know / context, don’t ask for information until it’s needed

Usability Testing

We usability tested 2 versions of the Alexa skill in the same sessions, in order to get more robust feedback on our live skill and a newly proposed demo. User testing was performed in order to narrow down which pieces of the experience needed to be revisited first. Usability testing helped us discover, prioritize and drive much-needed changes, as well as validate some of our recommendations in our demo skill. I worked with a UX Researcher, a PM, and Voice Engineer to create a test plan which included establishing testing goals, recruiting users, a moderator script, and ensuring we had a working prototype in a test environment. I alternated moderating these usability test sessions with our UX Researcher and recruited our study participants. 

Research Questions:

Qualitative

• How do users try to “open” the skill? 

• How do users describe destination points to Alexa?

• Do users feel they are given enough parking options to make a decision?

• Which parts of the skill most often cause frustration?

• Onboarding and permissions flow - do users get frustrated or confused?

• Can users understand the difference between bookable and non-bookable?

• Are users able to self-recover from errors? 

Quantitative

• How long does it take users to get through the flow?

• How many turns occur?