Tips for Creating Certifiable Example Phrases for Your Alexa Skill

Staff Writer Apr 14, 2017
Alexa Skills

As we’ve shared before, the majority of Alexa skills submitted for certification run into a handful of the same issues. Incorrect example phrases are the most common reason submitted skills fail certification. To help you avoid this issue, let’s recap the key requirements of example phrases and share some best practices.

What Are Example Phrases?

As part of your skill submission, you provide at least one example phrase under Publishing Information of the skill-creation workflow. To update example phrases, in the developer console, click on the skill name “[Skill Name]” IN DEV version > Distribution tab > Update 'Example Phrases' Field OR
click on + to add new example phrases.

These example phrases are displayed to customers under the “Try saying” section when they discover your skill. We like to think of this set of phrases as a starter guide that show people how to use your skill on their devices with Alexa. It’s also a great opportunity for you to highlight key features.

The basic structure on how example phrases invoke a skill is detailed in our Understanding How Users Invoke Custom Skills documentation, and is summarized as:

[Wake Word], [Launch Word] [Invocation Name] [Connecting Word] [Utterance]

Wake word: This is “Alexa” by default on Alexa devices, but can be adjusted by customers based on their preferences. Feel free to use “Alexa” as the wake word in your example phrases.

Launch word: As specified in our documentation, this includes a number of starting phrases including "open," "ask," "start," "launch," "begin," "talk to," "run," “load,” and more.

Invocation name: This is the invocation name you assigned to your skill under Skill Information within the skill-creation workflow.

Connecting word: These are words used to connect the launch word to utterances and include "and," "to," "for,” “when,” and more. For a full list, please see Understanding How Users Invoke Custom Skills. While this component could be omitted, including it would be more natural for consumers to understand and say.

Utterance: These are required and should be modeled based on the sample utterences within your interaction model.

In the example phrase “Alexa, ask TED Talks to play the latest talk”, “Alexa” is the wake word, “ask” is the launch word, “TED Talks” is the invocation name, “to” is the connecting word, and “play the latest talk” is the utterence.

And here's an example without a connecting word: “Alexa, ask My Pregnancy how far along I am," In this example, “Alexa” is the wake word, “ask” is the launch word, “My Pregnancy” is the invocation name, and “how far along I am” is the utterence.

Common Issues with Example Phrases

Here are some of the most common issues we are seeing with example phrases and their requirements:

  1. Missing components: Submitted example phrases are missing the correct invocation name or launch word in almost 10% of skill submissions. Let’s consider this example: "Alexa, ask to check my balance.” Without the invocation name specified after “ask” and before “to,” Alexa would not respond properly. Sometimes we also see the name of implemented intents here (as an incorrect substitute for invocation name).

    Here’s another example: "Alexa, Social Headline for what’s trending.” In this case, Alexa may not respond properly because the launch word is missing. We’re seeing issues with incorrect launch phrasing in almost 16% of submissions.

  2. Not based on sample utterances: Each example phrase should be modeled based on the sample utterances within your interaction model. For example, an example phrase of “Alexa, ask Tide Pooler when is high tide in Seattle” must have a corresponding utterance to provide a valid response


Copied to clipboard
<Intent Name> when is high tide in {City}

If the utterance does not exist, Alexa cannot map the example phrase to the correct intent. The skill would not know how to respond and the result would be a poor user experience. We’re seeing this issue in roughly 17% of skill submissions.

3.  Errant responses: The last callout here will be for you to ensure that each    example phrase provides a relevant response. In about 7% of skill submissions, we’re seeing issues with potentially confusing responses such as the followin

User: "Alexa, ask Home Cooking how to make pancakes."
Skill: "Welcome to Home Cooking. You can ask a question like, what's the recipe for pancakes. Now, what can I help you with?"

We hope you find these tips helpful in creating and submitting successful skills. If your invocation phrases are structured properly, are based on sample utterences, and provide relevant responses, you are more likely to move through the certification process quickly. Check out our previous blog post for additional tips on certification. 

Send Us Your Feedback

As usual, we’d love your feedback on certification. Please send us your thoughts at any time using our contact form.

Build a Skill, Get a Shirt

The Alexa Skills Kit (ASK) enables developers to build capabilities, called skills, for Alexa. ASK is a collection of self-service APIs, documentation, templates, and code samples that make it fast and easy for anyone to add skills to Alexa. 

Developers have built more than 10,000 skills with ASK. Explore the stories behind some of these innovations, then start building your own skill. Once you publish your skill, mark the occasion with a free, limited-edition Alexa dev shirt. Quantities are limited. 

Recommended Reading

Drive engagement by making your skill a part of customers' daily lives with the Alexa Routines Kit
Certification requirements for privacy policy URLs
Six tips for designing smooth Alexa shopping experiences to help grow revenue