What we are thinking about

This is an interview with Sean Fisher, author of over 5 live skills on the Alexa Skills Store including “Read the Book of Mormon This Year” and “Scriptures Every Day”.

Sean Fisher is an associate at premier technology consulting firm Pariveda Solutions. He has deep experience in web, mobile, and serverless technologies, and is excited about the new wave of voice-first interfaces. He lives in Houston and is a proud new father.  

Everyone of Sean’s skills are in the 95th percentile of retention, so VoiceLabs wanted to ask him how he did it, and share that with you!


Sean, help us breakdown how developers should think through retention, no matter the type of experience they’ve built.

In my opinion there are four main types of voice experiences, all of which have a varying “base” retention potential:

Utility skills. Smart-home skills, music, white-noise or “start a timer” type of functions. These connect with the physical world. The most important philosophies for these skills are simplicity and integration with “real life.”  A user may be chatting with somebody else, start a timer with Alexa, and resume the conversation with neither participant being distracted or even remarking on the interruption. In my opinion these are the types of skills that will have the highest retention – they are woven into the fabric of life.

Daily skills (like mine). People know from the get-go that they aspire to interact with your skill every day (if it passes their quality test). Examples include current events skills (including flash news briefings), quote-of-the-day skills, or goal-oriented skills like daily fitness routines. The most important philosophy for these skills is reliability and interface consistency.

Companion skills. The skill is just another channel for an already established brand, and the brand has existing communication channels outside of the voice-interface (email, text, tv, etc). The most important philosophy for these skills is integration with the brand’s other services. They need to integrate with the outside channels so well that people see the brand’s other properties and the skill as seamlessly complementary. Retention becomes a result of the brand’s complete strategy and effective consumer messaging.

Event-based skills. These are triggered based on a non-recurring or seldom reoccurring external stimulus. Examples include triggering a game when friends come over, listening to black Friday deals, or asking for a joke to pass a bored moment. The most important philosophy for these skills is delight. If a user is to remember to invoke your skill based on an event, than they have to be more than satisfied with your skill – they have to be delighted. In my opinion these are the skills for which retention is most difficult.

You mentioned a handful of philosophies when describing the various categories of experiences. Tell us more about what “reliability” and “simplicity” mean to you.

Code reliability. The instant your user hears the dreaded “There was a problem with the requested skill’s response” then they are out, for good. Monitoring and analytics are important parts of building a reliable experience.

Simple information requirements. Every additional piece of information the user needs to provide (typically using a special phrasing, I might add), requires them to build an increasingly complex mental interaction model. My skill by default requires no additional parameters/slots, and provides sensible options for small customizations. Any additional user effort translates to decreased retention.


Which philosophies have been most impactful for your own skills?

Focusing on single-purpose skills has been most impactful. Each skill should be purpose driven and extremely focused, leading to a simpler mental model for the user. There’s no UI affordance with voice-based interfaces. For example, my skill has only one custom intent (alongside the “meta” intents – help, stop, cancel, etc). Having single-purpose skills also helps guide you to an appropriate, simple, and memorable invocation name, which is the piece in the end that becomes your skill’s “identity” in the minds of the users.

The other philosophy is careful selection of a skill’s value to the target audience. In my case, the value the skill provides is extremely well aligned with my target audience’s habits. I know that my target audience is already trying to read the Book of Mormon or other scriptures every day – this skill just taps into their existing motivation and expands its reach. This value-to-need alignment is probably the largest contributor to a skill’s retention numbers. In my skills’ cases, I simply recognized the opportunity.


What’s one piece of advice that every developer can implement today to improve retention, with very little development time?

I would focus on injecting simple reminders. Before and after actually reading the content in several of my skills, Alexa will say, “here’s your daily reading…” or “That’s it for today, check back tomorrow!” Having short reminders will help users build a habit by increasing your skill’s mind-share.


Have questions, comments, feedback? Come ask Sean and the entire VoiceLabs community.