What we are thinking about


Share this on Twitter

At VoiceLabs we’re really enjoying the journey into voice-first interfaces, and we’re excited to announce a key new capability that came from a burning customer need: How do I know consumers are using my skill as I intended it to be used? Voice Pathing answers this question.

Compared to web and mobile apps, where users have the benefit of a screen, in voice applications it’s very easy for consumers to get confused. Once a user gets adrift, all hope is lost – that device in the home goes from a source of delight to a harbinger of annoyance and frustration.

Some consumer confusion / frustration points include:

  • What do I do next?
  • Did I say it right? And did Alexa understand me?
  • Why is Alexa asking me this question?
  • Why is Alexa talking for so long?
  • Why is Alexa making me feel bad about myself?

With the Amazon Echo, Alexa skills have a short amount of time to delight the user before the skills are forgotten.

A developer cannot predict all the issues their users will face learning a new interface, the only way to make the experience better is to understand and improve – hence Voice Pathing.

Voice Pathing shows the consumer experience and quickly bubbles up any confusing or missing functionality.

Each circle in the screenshot below represents a voice “intent” (as defined on the Alexa platform). These voice intents are the building blocks of any conversation.

Voice Pathing

How to Start Using Voice Pathing

To start using Voice Pathing, just install the VoiceLabs SDK into your Alexa skill. That’s it – we automatically build the behavioral flows and all the conversation threads, all you have to do is login to https://insights.voicelabs.co

What to Analyze First

Below, we’ll use a sample voice application to demonstrate how pathing can improve conversations.

As a developer, building great experiences starts with making sure that your users get through a ‘successful’ session. In this case, successful is defined by the user reaching a goal, and the developer creating a delightful conversation by walking the user to the objective, seamlessly.

Let’s hop into our sample voice application, Voice Pick’Em: a multiplayer game where users predict the outcome of today’s MLB games, and win prizes.  

Out of all the voice paths for Voice Pick’Em (in the above screenshot), two of them are successful experiences: predicting baseball scores, and checking the leaderboard. As can be seen in the screenshot, those are the second and third paths.

We always want our successful paths to be the most traveled, and developers should be on a constant mission to increase the percent of successful sessions.

Here is an example of a successful path:

Good Path

From looking at this path, we can see that users know what to do and what to say to get there.

Fixing the Bad

Let’s take the fourth path, which was not successful. Why did a user start to pick teams, then not make any selections?

Bad Path

By highlighting that path, we can start to dive into the exact conversation flow.

What caused the person to abandon? Let’s dig deeper into the Playball part of the conversation!

Here we are showing the raw SSML text that was tracked via the SDK, and what Alexa said to the user:

Screen Shot 2016-08-30 at 8.20.22 PM

Wow, what a mess! We start off by telling the user bad news. We then proceed to tell the user what they should have done better.

Twenty seconds in, we have bummed out this user and scolded them – probably not a recipe for success.

We finally get to why the user invoked the ‘Playball’ intent in the first place – to pick a game! However, this user already walked away angry and hurt.

From analyzing this flow, we can immediately start testing new conversations that put the good news up front, and give ‘helpful tips’ at the end of a session.

We just came up with a hypothesis and realized that we probably fell victim to the consumer thinking “Why is Alexa talking for so long?” and “Why is Alexa making me feel bad about myself?”

VoiceLabs promises you that Alexa is not mean spirited – developers just need a little love when building their experiences to make her be all that she can be!

The start of a journey

If you’re building a skill, get ready to iterate. An elegant conversational flow comes from trying out new styles, setting objectives, and monitoring how the user paths improve. Speech recognition is not perfect (yet), consumers will get confused or bored, and great ideas may come up short if not delivered in natural ways – that’s the fun of it!

With VoiceLabs’ Voice Pathing, you can quickly see the success your users are having, and make the experience better every day.

Here’s to great voice experiences!

The VoiceLabs team

Learn more about VoiceLabs