PROJECT SUMMARY
NBA developed an Alexa skill with a separate vendor a year prior to working with myself and my team. After reviewing the skill internally, we found many holes and gaps in the user experience that needed to be filled. NBA also wanted to incorporate schedule and score intents in addition to the radio intent that was already implemented. Our team worked closely with the NBA team to revamp the Alexa skill through consultation of the existing skills and the development of the new intents.
PROCESS
When developing a VUI, I have found it’s best to keep a crawl-walk-run mentality - build a solid base, refine and polish the base, release, and then build a little bit at a time as the skill ages. A voice skill should be an easy, natural, thoughtless interaction - which are the hardest user experiences to craft. With NBA we had to rebuild the base and incorporate two new intents. When considering the universe of content NBA owns we decided to focus on the two key intents that made sense for a VUI - schedule and scores. First, we came together to do a workshop and develop a list of intents the Skill would need. This gave us the direction we needed to build out the VUI flows. To button up the Skill, each intent needs an utterance library.
WORKSHOP - USE CASES & INTENTS
Together with NBA we worked towards defining a collection of use cases the Alexa skill could cover based on capabilities. The cases started out simple - What game is on tonight? When's the next Knicks game? From this, we were able to group these cases by what was being asked for. For schedule there were three main groups: one based on the schedule for a single team, for two teams, or a timeframe. Creating a skill is very much a balancing act between technical capabilities and user experience - more-so than mobile or desktop interfaces. Without a view, you have to direct your users through voice alone, and when something is broken, it's much harder to guide the user back to where they need to be.
After the workshop, we had a collection of use cases that defined our intents which we would continue to refine throughout the project.
VUI FLOWS
After defining a set list of intents that we would work towards, I created the voice interface flows that helped shape the experience. This included the new feature set of schedule and scores, as well as reworking some of the foundational intent flows that would help pad the overall experience. Cases like "what happens when Alexa can't understand a team name?" or "what if the user loses connection to wifi in the middle of an intent?" Thinking through these scenarios was part of building a solid base, and if we didn't do that, the more complex intents would fall flat.
Here are a couple examples of the VUI flows that were implemented in the Alexa Skill.
UTTERANCES & TESTING
To round out the Skill the next step is to build a collection of utterances for each intent. The list for each intent can vary quite a bit depending on complexity. For a simple intent like changing your time zone, the utterance list is fairly short. For an intent that is more complex, like asking for the schedule for specific teams, the list of utterances climbed well over 400. It’s important in a voice skill to make sure you have considered not only the combination of ways a user could ask for the schedule, but also dialects, shorthand and other nuances of human communication. Voice interfaces should be as conversational as possible, and to do that, it was important to build an utterance library that was as inclusive as possible.
After the team felt comfortable with the utterance library we had, we began testing. The whole team participated to make sure the updated Alexa Skill was ready to ship, and more importantly, be a valuable addition to the NBA brand .