Best Practices to Streamline the Alexa Skill Certification Process, Inspired By You
Every skill published to the Alexa Skills Store goes through a certification process, making sure that Alexa customers are presented with a delightful and engaging experience. Just as we are invested in providing a great customer experience, we are deeply invested in the developer experience for creating these engaging skills.
Every developer who submits a skill for certification has the opportunity to share feedback via a survey we send with their certification results. Based on the results of this survey, we’ve identified a few common certification pain points that are relatively easy to avoid. Addressing these issues upfront should lead to an improved certification experience for you.
1. Write an Informative and Accurate Skill Description
Your skill's description is likely one of the first things an Alexa customer will see when they encounter your skill for the first time. You will want to make sure that it is enticing and engaging, but most importantly, you will want to provide an accurate overview of the skill's core functionality.
Does the skill require the use of an additional service/account? Does the skill integrate with specific devices that customers need to have present? Are there any special instructions or limitations to this skill that customers should know about upfront? Answering questions like these within the skill description will help Alexa customers and the certification team understand the experiences that they will have when they interact with your skill. We encourage you to check your description against the Skill Description and Detail Page tests, which are valuable reference material to help you assess if you are providing the right amount of information.
2. Leverage Your Interaction Model for Your Example Phrases
To make sure your example phrases are understood by Alexa with high confidence, it is important to use the sample utterances defined in your skill’s interaction model. Ensure that the sample utterances support the invocation phrases, are free of typos, do not have grammatical errors, and contain all of the necessary components, which we outline below.
A standard structure you can follow when creating example phrases would be, “Alexa, ask [Invocation Name] to [Sample Utterance]” or “Alexa, open [Invocation Name] and [Sample Utterance].” Of course, you can make variations to these, but the wake word “Alexa,” the correct invocation name for the skill, and sample utterances taken directly from your interaction model are all required components for example phrases. The Alexa Developer Console will present some warnings inline if your example phrases are in an invalid format.
Something to keep in mind if you have multiple languages in your skill is that each language model will have an independent set of example phrases and must match that language model’s invocation name and sample utterances. Make sure to verify this for every locale your skill supports, prior to submitting your skill for certification.
For every skill you intend to publish to the Alexa Skills Store, review and test your example phrases. You can find a deeper review of example phrase best practices in this blog post.
3. Help the Customer Navigate Your Skill
It’s important to make sure that you provide help to customers and give them insight into the interactions that they can have with your skill. Instead of solely stating basic skill functionality and closing the session, the AMAZON.HelpIntent should be implemented to keep the session open and give customers the opportunity to make progress within your skill.
Here are three questions to keep in mind when providing help to Alexa users:
- Does the information provided help the customers navigate the skill’s functionality?
- Does the help interaction end in a question, requesting the user to take an action?
- Does the help interaction result in an open session so that the user can continue interacting with the skill?
If your answer to any of the questions above is “no,” you’ll want to update the AMAZON.HelpIntent accordingly before submission.
Visit Providing Help for more help implementing the AMAZON.HelpIntent.
4. Provide Access to the Certification Team
If your skill has account linking, be sure to add a valid test account’s login credentials to the Testing Instructions section of the Distribution tab. The certification team will use these credentials to review the complete ins and outs of your skill based on the certification requirements here and here. If the credentials provided have expired or are invalid, the team will not be able to access your skill and complete the certification process.
A best practice for the customer experience is to mention account linking requirements in the skill description.
5. Set Up Discoverable Devices with Regional Endpoints
If your skill requires specific devices to function to streamline the certification process, the certification team will need to verify that the devices are successfully integrated. For this process, it is important to provide test account credentials that are connected to the devices, as mentioned above, and that the devices are online and discoverable.
Geographic-specific endpoints are a must when developing a smart home skill to ensure fast discoverability and quicker skill responses. The skill’s language model must be associated with the correct lambda endpoint. The list of skill languages and the applicable endpoints may be found on the “Configure the Smart Home Service Endpoint” section of the Steps to Build a Smart Home Skill page.
By following these best practices, you’ll be able to avoid some common issues that other developers have encountered during skill certification and provide engaging voice experiences for customers. In the meantime, please continue to share your feedback on the certification process and skill building in general via our surveys and the Alexa developer forums.