Tips for Planning an Amazon Alexa Skill

Written by Andrea Moretti, Technology Lead at Apply Digital.

The Value of Voice-First Products

Consumers are devouring voice-first access to services and apps as a convenience they never knew they needed. As a result, many companies are eagerly developing their own branded voice-first Alexa skills and evaluating how voice apps might fit in their product mix to extend their reach and provide new entry points to their products.

Voice-first products resonate with customers because they provide natural, hands-free, frictionless interaction with the devices and services that make up their digital world. People are becoming accustomed to this new mode of interaction quite quickly, and products with well designed voice-activated abilities are starting to have a greater edge over competitors.

Your Two-Minute Alexa Development Primer

Alexa skills development is similar to that of any other software. The main difference is understanding its domain-specific terminology.

Skills

An Alexa skill is a voice-first app running on the Alexa platform. It is like an app, but skills are not called apps to avoid confusion with Amazon’s own Alexa control smartphone app that supplements Alexa skills.

Skills come in three types: custom skills, smart home skills and flash briefing skills. The latter two are specialized with minimal user control, so in the discussion below we’ll focus on custom skills.

Intents, Utterances and Slots

Think of an intent as one of a set of specific actions in your skill, much like the individual functions in a conventional computer program. Each intent is contains a set of phrases expected from a user to invoke the intent or perform some action. Optionally, one or more slots in each intent provide a way to extract typed parameters from an utterance.

Utterances typically require the most work in both planning and development, since customers expect your skill to respond flexibly to a variety of synonymous phrases, which could number in the hundreds.

Slots convert specific words from an utterance into typed data such as dates, activities or time that help the intent to fully complete a requested action. Alexa provides a large library of built-in slot types, but developers can create custom slots as well. Custom slots may also be used in a limited way to provide wildcard functionality in utterances.

Invocation Name

An Alexa custom skill is accessed via its invocation name. Since the name of your skill may clash with other skill names, it must follow strict rules in its creation. It may occur almost anywhere in an utterance. This name activates your skill in the same way as tapping an app icon in a visual interface.

To learn more about Alexa’s terminology, check out our blog article: Speaking Alexa’s Language: Terminology for Building an Alexa Skill

Tricks and Tips for Creating Your Alexa Skill

Planning Tips

Like any type of consumer app, you need an end-to-end planning approach that carefully considers marketing goals, requirements, development effort and support for scaling and performance along with measurements of how and how often customers use the skill.

Keep in mind always that the interface operation of an Alexa skill is distinctly different than that of screen-based apps on devices with keyboards or mice. For example, the user must essentially hold in memory the interface of a skill rather than being able to visually scan it.

Misunderstanding such aspects often leads to an undue burden on skill users. Following best practices for voice UX and tracking usage patterns from first contact through every command-response path will shine light on the most common interface flaws.

Your Brand Persona

New Alexa customers adapt to Alexa-enabled products in a manner similar to how they interact with real people. Currently, you cannot change the voice or play background audio. However, that does not mean you cannot add some personality to your skill via the specific verbiage of its responses or in the way, say, options are presented. Such customizations can set your skill apart from the herd and enhance a positive reception of your brand or product if appropriate.

Development Tips

Conversations

The ability to engage in multi-turn conversations is a new Alexa capability. It is an opt-in feature on the Echo, Echo Dot and Echo Spot called “follow-up mode,” which tells Alexa to standby for follow-up questions without the user needing to repeat “Alexa, …” for their next request.

This feature is provided by the platform’s Dialog interface. It may be used to collect additional parameters for a search or multi-factor problem diagnosis. Additionally, dialog management capabilities maintain multiple conversation threads as users switch between topic contexts or other skills.

Utterance Tracking

When choosing your utterances, try to include as any variations as you can think of. Remember: it’s better to have too many than too little. The Alexa Dashboard lets you visualize how well any intent’s utterances are working. It displays which ones are used, which fail most often and so on. Utterance sets are likely to be a primary focus of your future skill updates, can you can fine tune your utterances over time

Cards

Typically, a visual-first app is developed alongside the voice-first Alexa skill. These provide a means for manipulating skill settings or offering additional information to users such as images, videos, extensive text or lists. Let Cards enhance the skill versus overlapping with it.

Currently, Cards are pretty limited. The Card layouts for mobile apps are predefined, and you don’t have very much control over the look and feel. You also can’t include links to other parts of your mobile app (deep linking). There are a few more options for Card layouts on Amazon devices that have a screen (ie. the Echo Show), but we’re hoping that more options become available for mobile.

Time Zone Limitations

For skills that need the local time, a custom or third-party geocode service is required since Alexa exposes a user’s location (assuming the appropriate permissions are granted) but not their specific time zone. This means that it’s much harder to respond with sentences like “Starts at 7:30 PM” than it is to answer “20 minutes from now” as Alexa, by default, doesn’t know a specific user’s timezone.

Now, the tricky part is that you cannot use Google’s geocoding service because the Google terms and conditions only allow you to use the geolocation API in conjunction with a Google Map, which is obviously not suitable for a voice interface, so we used HERE, which is an API originally created by NOKIA.

Push Notifications

Recently, Amazon added the ability for skills to send out notifications to Alexa-enabled products. This is a strictly opt-in feature from the user’s side using the Alexa App settings feature. The presence of notifications may flash a different color on LED-equipped Alexa devices or send an audible alert, but when to listen to notifications is completely under user control.

Skill Publication

In contrast to other mobile app submissions that let you choose the exact time your app goes live after approval, Alexa skills are published immediately after approval, so keep that in mind when trying to time your skill’s release.

Staging Environments

Alexa doesn’t have a concept of environments, like a development, staging or production environment. You need a development environment to build in, a staging version for QA and then a production environment for the version you will ultimately publish, but with Alexa, you basically have to manually duplicate your skills and manage three different versions of the skill in these different environments.

You also aren’t able to transfer a skill from one developer account to another. For our client projects, we had to first set up the production environment for the skill on our account, but to transfer the skill to our client’s account, we had to duplicate the skill.

The Scramble for Voice Market Share Is On

Right now, Amazon’s Alexa holds about three-quarters of the market share for prevailing voice technology products along with meteoric growth in the number of products in its Alexa Skill Store.

Over the 2017 Christmas season, their Echo products were the top sellers in their vast catalog of products. Continued explosive growth is supported by Gartner’s estimate that this year 30 percent of smart machine interactions will be voice-based.

Imagine in the very near future a user asking their voice-enabled device about your product and receiving no response or, worse, receiving a response from a different company that has already claimed your brand.

Partnering with Voice UX Experts

Apply Digital provides companies with a digital product team focused on AI and machine learning products. Apply’s voice-first UX and “smart products” talent and expertise are invaluable in determining precisely where to place a voice-first branded presence in your company’s strategic goals.

If you’re interested in creating a voice-first product or an Alexa Skill, speak to us at applydigital.co.

Apply Digital is a Digital Product Studio. Apply partners with organizations to concept, design, build and manage digital products and services. By leveraging artificial intelligence and machine learning, they create “smart” digital products that achieve business goals and provide utility to end users. Apply has built a world-class team of researchers, UX specialists, designers, strategists, developers and data scientists to create compelling, ROI driven products. They have offices in Vancouver, BC and Toronto, ON. To learn more about Apply Digital, please visit applydigital.co.