Categories
Gabi's Blog posts

Alexa SEO

Alexa SEO is the effort to increase the number of organic installs of your Alexa Skill, making it easy for the correct audience to find and enable your Skill.

In this post, I share both basic and advanced Alexa SEO techniques to increase the number of organic installs of your Alexa Skill. This post was originally meant to be kept internal to our startup (Chicisimo), but then we thought it could be very helpful to some teams out there looking for non-paid acquisition.

At Chicisimo, we discovered that our learning method for ASO growth can be successfully applied to Alexa SEO (also known as Alexa Skills Store Optimization). At Chicisimo, most of our 5 millions iOS and Android installs come from ASO (or word of mouth from people who found us via ASO).


Our previous method to succeed in ASO

First is first. The basic receipt that made us successful in ASO are iterations and documentation:

  • Iterations. We’ve been iterating at least once a week during 4 years. Most of our new builds sent to the App Store had an ASO test;
  • Documentation. Each ASO test has been documented.
    • A spreadsheet keeps track of evolution and impact of each ASO test;
    • And on Github, we’ve been documenting the tests performed and the lessons learnt.

These 2 simple aspects developed into a method.

Chicisimo’s Alexa Skill: An in-bedroom fashion stylist. More info here.

Introduction to Alexa SEO

The biggest challenge faced today by Skill developers is the discovery of their Skill by its potential customers. Skill developers and publishers need to understand what different mechanisms exist to be discovered.

This post explains how to get your Alexa Skill discovered organically. It is a complete guide to SEO for Alexa Skills, what is called Alexa Skills Store Optimization in Amazon’s Skills Store.


What is Alexa SEO

Alexa SEO is the effort to increase the number of organic installs of your Alexa Skill, making it easy for the correct audience to find and enable your Skill.

Amazon Skills Store is kind of similar to the App Store a few years ago.

Home page of the Alexa Skills Store

Alexa SEO – Skills Store Optimization is a new discipline with little documentation and lots of iterations. Amazon Alexa Skills team is learning how developers and customers discover the best Alexa skills, and they are iterating fast to facilitate skills discovery. Alexa Skills and new hardware will continue conquering our lives, so being able to discover great skills is a key process, also for Amazon.


Basic Skills Store terminology

Amazon Alexa apps are called Skills. Installing is called “enabling”. Thus, “Enable a skill” is the equivalent to “Install an app” in iOS or Android. The Analytics tool in the App Developer Console talks about “Activations”. Skills are not actually installed, they live in the cloud.

Customers enable skills in the Alexa App, from Amazon.com/skills, or via voice by saying, “Alexa, enable [skill]. There are important exceptions to this last option, which we’ll cover below.

“Invocation name” is the set of words that customers use to invoke and interact with your skill.

If you want to find further content on the SSO topic, you can search Google by “Alexa SEO”, or “Voice SEO” which is quite different but you’ll still find content.


Build the Best Alexa Skill

The number one driver of organic growth is creating a great experience for first time users and recurring ones. I realize this might sound off-topic when talking about Alexa SEO, or idealistic, but it could not be truer. The best skills will onboard people correctly and will drive engagement, they will get word of mouth, the Alexa team will notice them and will featured them. All these factors will add up, and the Skill position in all types of rankings will greatly increase.

Nobody has ever built a great skill without product iteration and a good understanding of the customer, so focus on building a solid learning process. Lessons from other products apply here: ship fast, ship often, focus on the process, learn. Best Alexa Skills take lots of iterations.

The onboarding of new users is a critical process for your Skill success. It will lead to retention, word-of-mouth, and will make your Skill unstoppable. Obsess over the onboarding, which we define as the process by which new activations find the value of the Skill as soon as possible, and before we lose them. You can learn about our iteration process by reading this post on How we grew from 0 to 4 million women on our fashion app, with a vertical machine learning approach.

In the case of our Skill, Outfit Ideas, we articulated to ourselves what needed to happen in the Skill onboarding. It was something like: If people don’t do [action] during their first x minutes in their first session, they will not open the Skill again. So we need to iterate the Skill onboarding experience to make that happen. We also run user-tests with different types of people, and observe how they react to the retention lever. BTW, although it is not the scope of this post, we identify retention levers using behavioral cohorts.


Sources of Skills discovery

There is still no formal research into the different sources to discover Skills. Leaving PR and paid acquisition aside, this is my guess on the top Skill discovery mechanisms:

  • Search and browsing – Alexa SEO
  • CanFulfillIntentRequest
  • Recommendations from friends
  • Featured Skills
  • Google

Skills Store SEO – Visibility & Conversion (search and browsing)

Two growth objectives matter in Alexa Skills Store Optimization:

  • Increasing visibility;
  • Increasing conversion.

1. Increasing visibility. You want your potential customers to find your Skill, similar to your objective in traditional ASO or SEO efforts.

The main factors for a successful Alexa SEO strategy are the same as in App Store Optimization, but adapted to the conversational context: keyword research, implementation, analysis of results, and iteration, iteration, iteration… You want to understand how your customers search for your Skill, and Skills like yours. By understanding how people think and describe their needs related to your Skill and competing Skills, you’ll have a great first Alexa SEO step. You can start the learning by using the same tactics you’d use for ASO. Build a first thesis by finding recommended keywords in the Apple Search Ads or Adwords keyword planner, and adapting them to the voice context. Then, ask your first customers what keywords they used to find your Skill or related skills; and then, iterate heavily.

What is most relevant here is the process you follow, we’ll talk about that below. Regular metadata iterations with careful analysis of results will give you control and growth. New releases of your Skill iteration should not just come when you want to update or AB your product, but also when you are seeking to test new SSO hypothesis.

The Analytics tool in the App Developer Console is still in its infancy, and does not provide much data. It does not provide, for example, activation data per source (direct activations, activations via the search box, or via Skills Store features). This means that you’ll have to measure the impact on total activations, and learn from there.

The main drivers of organic activation in the Skills store are: company name, skills name and descriptors. You can read about each of them below.

2. Increasing conversion. You want visitors to your Skills page to fall in love with your skill, and enable it.

The main drivers of conversion are: the Skill name, the Skill Icon, the rating and reviews your Skill receives, and utterances that you feature (called Alexa Speech Bubbles). The Description is also kind of relevant, although customers won’t read more than a few lines.

Again, the Analytics tool in the App Developer Console will not give you a lot of insight into your iterations, so you will have to measure the impact on total activations as in the case above. You can follow the same tactics as many do in the case of the App Store, mimic your Skill page, and AB different options in a browser with real users, even though this is not perfect. You can also do a before vs after AB, but the validity of the results will depend on the amount of visitors you have – the statistical significance of the data.


There are 6 fields to include keywords in Alexa SEO

In the Skills Store, there are 6 fields to include keywords: Developer name or company name, Skill public name, One Sentence Description, Detailed Description, Example phrases, and the Keyword field.

Developer name or company name. This is the field where keywords have the highest weight.

You need to introduce this name in your Company profile when you create an Amazon Developer account. You cannot change the name after you create your account, although I guess that if you change the name of incorporation of your company, there is process to change your dev name too (as in the App Store).

Skill public name. “The name of the skill that will be displayed to customers in the Alexa app. It can be different from your invocation name and must be between 2-50 characters”.

This is very important for three reasons:

  • Keywords in the Skill title rank the highest compared to other fields, excluding the Developer name;
  • Skill name builds the URL of your Skill. It will end up being indexed by Google, and that will bring new installs from people searching on Google. Remember that, in the case of apps, people search for them both on the Stores and on Google. In the case of the Skill Would you rather for family the Skill name is included in the URL, while in the case of Archer Trivia both the company and Skill name are included. There are other examples that differ from those two. We’ll monitor how iterations affect URL. Of course, the key identifier in the URL are not the keywords, which can be removed and the link will still work, but the Skill ID (same happens in ASO). In our case, our most basic URL is https://www.amazon.com/dp/B07W6RSJG4/ but the default URL that Alexa shows includes the name (our current name) https://www.amazon.com/Chicisimo-Outfits-Closet-Fashion-Outfit/dp/B07W6RSJG4/ and depending on how you get to that URL, there is a suffix of elements indicating the query ID etc (example 1, 2, 3 and 4). I have not seen a country or language identifier in the URL – Searching for “Outfit ideas, by Chicisimo” on Google does not return the Alexa link a few days after launching the skill and no building links to the Skills page. On Aug 20th, 2019 we tested linking to our Alexa page from several sources, and a few days later the page got indexed! There you have another area to work on;
  • Keywords in the Skill name do affect SEO, same as ASO App Packs;
  • And IMO most important reason: Skill name is one of the most visible elements people see when they search for Skills, and it’s a big driver of conversion. I liked how they describe it in their Tips on Promoting your Alexa Skill: give your skill a clear and intuitive name that speaks to its purpose and value. The name should set clear expectations on what your skill can do and what the customer experience will be like. Your skill name should also be easy to remember so customers can recall it when they are ready to activate your skill.

The Skill name we chose is “Outfit ideas, by Chicisimo”. Outfit ideas is a popular keyword both on search engines and app stores. It also clearly sets expectations, and it’s easy to remember. We’ve added by Chicisimo because many people who know us might expect our name to be there, but we realize it’s not easy to pronounce. Well, we’ll see, this is all work in progress. Our invocation name is Outfit ideas. It’s important to note here that you cannot change the invocation name after a skill is certified and published.

One Sentence Description. “A quick at-a-glance sentence that describes the skill or what customers can do with it. This will display in the skill list in the Alexa App.” Maximum length is 160 characters.

Detailed Description. “A more comprehensive description of this skill. Include information about any prerequisites like hardware or account requirements and detailed steps for the customer to get started. This description displays to customers on the skill detail card in the Alexa app. For Flash Briefing skills, list the feeds offered within the skill”. Maximum length is 4,000 characters.

Who reads a 4k character description? Nobody! Do keywords here help rank in the Skills Store. No. So from an SSO point of view, why is this relevant? Well… Google. Google does index the short and long description in the Skills Store, so use it wisely. I would not invest a minute into this in your first SSO iterations, but I’d take into account in your future iterations. You can learnt about this yourself by taking a piece of text from the description of any skill, and searching for it on the Skills Store and Google. Here’s Google result and the Skills Store result here.

Example phrases. “Sample utterances that will appear on the skill detail card in the Alexa App. These phrases will help users get started and enable them to access your skill’s core functionality”. You can include up to three utterances, the first one being an example on how to open the Skill. In our case, our 3 utterances in the first iteration where: “Alexa, open outfit ideas”, “Give me outfit ideas for today” and “How can I wear my red dress”.

Keyword field. You can include 30 keywords, and there is no character limit. Many people often ask if you can use competitors names in this or other fields. What they tell you is that you have to respect intellectual property. The implementation of this rule is normally flexible and depends on the reviewer.

These keywords are not seen by customers. This is what Alexa says about this field: “Simple search words that relate to or describe this skill. This helps customers find the skill quickly and easily. Use spaces or commas in between each search term”.

What’s new? “Use this field to let customers know what new feature(s) this version introduces or what issues are resolved with this version. If this is your first version you can leave this field blank”. Maximum length is 2,000 characters. In the App Store, this field is not taken into account for rankings in the Skills Store. It’s indexed by Google, but only the last “What’s new” so it has zero value, at least for teams with regular skill updates. It’s the same in the Skills Store.

Category. You need to select one Category.


There are 3 Alexa Speech Bubbles

The Alexa Speech Bubbles are one of the most noticeable visual aspects for the visitors to your Skill page. A very quick and visual way to give customers examples of what they can ask your Alexa Skill. The Alexa Speech Bubbles are a graphic way to illustrate examples of Alexa utterances that trigger Alexa actions. You can see details here.


Ratings and reviews

Ratings and reviews written in the skills store in one country, are visible to users from other countries, following the Play Store model and not the App Store model. Hiding my IP using Tor, I visited the skill store via Austria, Spain, Italy, the US and other countries, and all of them showed the same reviews and ratings.


Different languages

ASO has proven that having your app in different countries/languages is an opportunity for growth in the App Store and Google. Skills Store optimization also benefits from this: using different countries to heavily rank for a certain word is something you should consider at some point. This is a great place to do a lot of testing.


Name-free Interactions with CanFulfillIntentRequest

According to Amazon, millions of customers use skills every month without using the invocation name. Name-free Interactions are those cases when a customer asks something to Alexa, without mentioning a skill name. Your potential customer has a need that you can help her with, but she does not know your skill, or simply she doesn’t invoke it, for whatever reasons.

An example would be “Alexa, give me outfit ideas to go the the office”, in this case Alexa looks for skills that might fulfill the request. Alexa determines the best choice among eligible skills and hands the request to the skill.

There is some friction in the Skills Store, because it order for a user to be able to use a skill, she’s had to enable that skill previously. Remember what we said above, “enabling a skill” is like “installing an app”. Name-free Interactions (sometimes called Nameless Invocations), shipped in May 30 2018, aim at solving this huge problem. Amazon tests and rotates Skills suggestions to see which skills customers are responding to and give more developers visibility. Suggestions are delivered using two techniques: a neural model architecture (Shortlister), and a hypothesis reranking network (HypRank).


Skill Auto Enablement

Alexa automatically enables thousands of the most popular skills, so your objective as an skill publisher is to get to this list. Auto-enablement means that customers can use a skill immediately when they say, “Alexa, open [skill]”, without having to enable it previously.

Getting to this list of auto-enabled skills requires you to build a great experience, and which is easy to enjoy – an easy onboarding.

Some of the criteria they use to choose auto-enabled skills include customer satisfaction, such as number of customers, customer engagement, and ratings, as mentioned above. When a customer invokes one of these skills, whether through direct or name-free interaction, they can immediately engage with the skill. For other skills, including those that have a mature rating or extra steps like signing in through account linking or parental permission, customers will be directed to use alexa.amazon.com or the Alexa App to choose and configure skills to enable before launching. Beyond ensuring you have an engaging skill, there are things you can do to give the system explicit signals, like adding accurate descriptions and keywords in the skill metadata, selecting the correct category, and implementing CanFullfillIntentRequest.


Build a learning process

Throughout this post, I’ve focused on iterating several times. Part of building an Alexa Skill and actually any product, is defining how the team learns. One of the most interesting aspects of building an Alexa Skill is how fast you learn in this territory, which help us improve the product significantly. When we’ve obtained the most relevant learnings, it’s always been because we’ve focused on two aspects: how people relate to the problem, and how people relate to the product. The question always is: how can we learn… sooner?

We have a process to learn about Store Optimization, which is based on the analysis of continuous iterations of the metadata and other factors that might influence Store Optimization. The result of this process is new learnings in each iteration.


The Skills Store review process

  • Time to review: In one email they sent us, they said they aim to complete Skill reviews in 5 days max: “Please note that we aim to complete testing and provide feedback within five business days of developers submitting their skill”;
  • In our case, our first iteration of the Skill was rejected 5 times, but the good news is that each rejection took them less than 24 hours after submission. The process was fast, and fast is good;
  • The reasons for rejection were always different, so we were surprised not to have one rejection including all issues. Would’ve speeded up the process;
  • One time, they rejected the Alexa Skill because they could not reproduce an utterance which we had included as an example in the Description field. Big fail on us. Anyway, this proves that Skill reviewers do read the Description;
  • Although we feel reasons for rejection were different, they thought otherwise. We got a couple of emails saying “We observed the skill is rejected multiple times for the same reason(s). In case you require assistance to fix the issue, please use the contact form here and we will help you in fixing the issue as fast as possible”;
  • We once contacted the Skill review team through the above form, with a specific question. Their response was standard. Probably not generated by a machine but a human, but I guess it was simply a standard response selected from different standard responses they have. They did not even attempt to respond to our specific question. For the sake of comparison, in the case of Apple reviewers, they either do not reply to your questions, or they give you a very specific response.

Reasons for the Skill rejection

This is a copy paste of a few reasons for rejection we received from the the Alexa Skills Team. Subject of the email is always “Certification feedback for your Alexa skill”:

  1. The skill does not deliver a LinkAccount home card to the Alexa app, when an intent requiring account linking is invoked by users who have not yet linked an account.
  2. Your Alexa Presentation Language (APL) skill has failed certification as it displays an “Invalid Directive” error message in the home card section of the Alexa app for devices without display (Echo, Echo Dot). This is because your skill is sending APL documents to devices that do not support APL.
  3. The core functionality of the skill could not be completed using the customer facing information. Make sure that the skill’s description and prompts make navigation of the skill easy and the core functionality clear to users.
  4. When invoking the skill with one or more of the intent(s) or sample utterances, the skill’s response is irrelevant to the request or contains an error.
  5. When attempting to invoke the skill using an example present in the example phrases or the skill’s description, the skill returns a non- contextualized response or error. All customer facing example phrases must function without error since these are interactions that users are most likely to try.

Useful links

  • Alexa Skills Kit here
  • Submission Checklist with Certification Requirements here
  • Amazon-moderated forums for Alexa Skill developers here
  • Alexa Skills Kit weekly office hours on Twitch TV, here
  • Amazon Developer Blogs, Showing posts tagged with Alexa, here
  • Form to contact the Alexa Skill reviewers here
  • Alexa Skills Store here

SEO for Google Actions

Another aspect that we’ve liked is how to make a Google Action easy to find with implicit invocations. We’ve learnt how doing a good job on Skill Store Optimization or Google Actions has a great impact on your overall SEO strategy.

Final thoughts

In my opinion, Alexa Skills are going to be part of the everyday life of many many people, at some point. I believe that we can predict an evolution of Amazon Alexa Skills Store similar to the evolution we’ve seen with Apple’s App Store during the last decade.

As with most things in growth or product, developing a solid iteration method is the key to successful growth. I hope the above info was helpful, please let me know if I can help with anything.

Have fun!