The world of voice skills is changing, and it is becoming more and more difficult for developers and companies to launch successful Alexa skills. From full development to no-code development, things aren’t getting easier when introducing these skills into the market and receiving positive reviews. 

Amazon announced that 100,000 unique Alexa skills had been created worldwide at the end of September. 

• The largest of those markets is the U.S., which has 70,729 skills alone. 
• The U.S. market is followed by other English-speaking countries, such as the UK (36,341), India (32,879), Canada (25,950), and Australia (24,651).
  

Creating an Alexa skill has never been easier. However, 2020 is considered the year with the lowest number of newly added Alexa skills. This year we are facing the lowest number of launched skills since 2016. 

The growth rate during 2019 was more than five times lower than that of 2018. Ironically, the number of Alexa skills users is growing, while creating an Alexa skill is shrinking. Why is this happening? We analyzed more than 13k bad Alexa skills and an overall 30k Alexa skills reviews to answer this question.  

Key Findings on bad Alexa skills

From more than 30 000 skills analyzed, we found that only around 37% of them have reviews. Among the skills with reviews, only 19% of them had positive reviews. Considering this, we might conclude that now more than ever, the focus will be on quality rather than quantity when creating an Alexa skill. We have decided to share with you our main key findings as per below:

• Three main problems users face when using an Alexa skill 
• Most popular categories with bad reviews 
• What is a bad Alexa skill? 

It is challenging to have a magical formula when it comes to creating an Alexa skill. Through experience and user feedback, we aim to step into the improvement process and learn new things. We have spent time understanding how our study would help you in the voice world and are open to added value. Our CEO and co-founder, Juxhin, has also added a personal touch to this study, which you might find helpful. At the end of this article, you will have a section on analyzing your Alexa skills using Ipervox. 

“Take time to define the problem you’re trying to solve through your skill. The general rule of thumb in creating a good user experience is don’t just build for the sake of building. Understanding the market of users who would use this skill and designing Alexa to speak in a level of formality and tone familiar to the user essential,” 

Diana Lee, Conversational AI Designer at Wizeline

Three main problems users face when using Alexa skills 

In our study of +13K bad Alexa skills reviews, the three main problems users faced when using an Alexa skill were as follows:

• 35.56% said that the alexa skill does not work
• 14.6% said that it is a bad alexa skill
• 6.54% considered the alexa skill to be useless

Why should you consider a “does not work” review when creating an Alexa skill?

The problem with voice apps is that users can’t see what they do. As a result, the first sessions tend to be exploratory: users want to test the Alexa skill’s limits by pushing all the buttons, pose weird requests, and test recovery from giving misinformation. That is why you should consider testing the design a lot during the Alexa skill development phase. When users are in their testing phase of your skill, they can create a first bad impression resulting in a bad Alexa skill review. 

% of [DOES NOT WORK] Alexa skills per each category

We can imagine Alexa skill stores as a search engine, like Google. Having skills with no ratings negatively impacts the Alexa skill store’s ranking and user acquisition. According to Chatbot, 40% of users discover a new skill through the store. Lack of reviews or bad reviews will cause the skill to disappear from the first page of the Alexa Skill Store and decrease its usage. Therefore, when creating an Alexa skill, it is essential to consider a strategy to increase positive reviews. We analyzed the percentage of “does not work” reviews per each category and summarized the findings in the table below. 

Category	Percentage
Business & Finance	35.29%
Connected Car	44.26%
Education & Reference	31.68%
Food & Drink	22.44%
Games & Trivia	23.32%
Health & Fitness	30.46%
Kids	37.37%
Lifestyle	35.00%
Local	27.27%
Movies & TV	26.01%
Music & Audio	41.43%
News	15.42%
Novelty & Humor	18.77%
Productivity	40.35%
Shopping	36.90%
Smart Home	41.36%
Social	34.46%
Sports	53.52%
Travel & Transportation	34.74%
Utilities	40.96%
Weather	15.22%

As seen from the data above, around 50% of Alexa skill categories have more than 35% of their bad reviews related to skill not working at all. Considering this, Amazon raised awareness towards Alexa skills developers and asked them to invest more in their Alexa skills quality than quantity. 

Every Alexa skill should have a well-thought interaction model that defines the words and phrases that users will say to Alexa to make the skill do what they want. Failing to consider this will increase user dissatisfaction. 

Most popular categories with bad reviews 

Our study also tried to understand which were the most popular categories with bad reviews among twenty-one included. This means the categories with the highest number of reviewed Alexa skills. We understood that five categories were leading the results as per below: 

• Connected car -592 skills were reviewed within the connected car category.
• Food and drink -664 skills were reviewed within the food and drink category.
• Utilities -664 skills were reviewed within utility category.
• Health and Fitness -929 skills were reviewed within the health and fitness category.
• Start Home -5169 skills were reviewed within the smart home category.

What is the future of smart home Alexa skills?

One thing is for sure; they need to become more intelligent and user friendly. Alexa is continually becoming smarter. A number of new upgrades were announced for Alexa, including a ‘memory’ feature, which enables the assistant to remember essential dates automatically in the background. An additional feature called ‘context carry over’ allows an Alexa user to ask one question after another without repeating ‘Alexa’ or ‘Amazon’ between queries. However, what we have seen from the reviews is that Alexa skills launched lack both “memory” features and “context carry over” features. 

What is considered a bad skill?

We have decided to explain a little bit further what is considered a bad skill from the most popular categories of the study. You can understand the users’ issues related to smart homes, connected cars, food & drink, utilities, and health & fitness voice apps. 

Smart home

From 579 reviews considering the Alexa skills as bad, here is what we found: 

One of the biggest concerns among Alexa users is the linking of multiple devices together. However, even after installing all the necessary apps to link these accounts and new discovering devices, Alexa does not control them. Redundancy is a significant problem: it is challenging to ask simple questions and simply get back simple answers. Choosing the right Voice Utility Spectrum when creating an Alexa skill can prevent bad reviews.  

Connected cars

From 83 reviews considering the Alexa skills as bad, here is what we found:

Connected cars Alexa skills cannot usually connect with the vehicle or maintain that connection when they do. Besides, users say that it takes a lot of time for the Alexa skill to send the signal and perform the asked task. When it comes to the output, Alexa skills reviewed within this category lack the proper instruction and therefore don’t answer to the exact same thing questioned. 

Food and Drink

From 79 reviews considering the Alexa skills as bad, here is what we found:

It is difficult to connect the Alexa skill with the echo device. Users expect Alexa skills under the food and drink category to be easy, but they turn out to be very complicated according to their ratings and feedback. The interaction with the Alexa skill is low since users cannot add new content to the skill (such as a receipt) and cannot save their preferred content. Failing to test the skill first causes reviews such as having a skill with no available samples to be shown to the user with insufficient information and name pronunciation problems. 

Utilities 

From 93 reviews considering the Alexa skills as bad, here is what we found:

Alexa skills within this category do not understand the given instructions from users. They perform even worse after the updates by not giving the promised output to their users. Often skills within this category fail to work, and users face bad experiences when deleting them. Not being able to understand what users are asking, the Alexa skill interrupts them. The UI user interface is not good and not user friendly. Besides, Alexa skill is asking users to use their phone applications, making them worthless to use.  

Health and Fitness

From 75 reviews considering the Alexa skills as bad, here is what we found:

Users say they will need to access the Alexa skill in another moment of alexa skill development, as for now, it does not respond following the input given. The Alexa skill is understanding less than the regular Alexa does. Furthermore, the content of the alexa skill is not related to what was promised. Alexa skill is either speaking too fast or is noisy, making the user feel uncomfortable. 

What makes a good alexa skill, then? 

According to Juxhin Rradhima, Ipervox CEO, The general rule to keep in mind is that a skill must make life easier for people who use it. It does so by lowering as much as possible the cognitive effort required to carry out that activity.

When creating an Alexa skill, it is crucial to develop a good Voice Design to rely on regular conversations to make the interactions as natural as possible. To do this, we must:

• Write on a piece of paper a sample conversation (script) that the user might have with the skill and recite it repeatedly with a person to see if that conversation makes sense and flows well.

• Start with a simple and appropriate script, which starts and ends with completing the required action and where the conversation between the user and the skill can be entirely predictable. 

• Find and test different directions in which the conversation can go, adding nuances, variations, etc.

• Therefore, structure an optimal Voice Design, which facilitates people’s lives and responds quickly to what the user requires, despite the various interaction possibilities.

Alexa Skills – get it analyzed with Ipervox.

The world is evolving towards voice more and more every day. Many “no-code” skill development opportunities out there make it easier and more comfortable to enter this frictionless world. Lately, Amazon has also included the pricing per amazon Alexa skill, which introduces us to a future with no more free skills. Should we consider this as good news? Can you develop an Alexa skill with no code experience and have the chance to get revenues from it? It depends. If the Alexa skills development does not result in a useful skill for users, it might fail to be reviewed or have bad reviews. 

At Ipervox, we have created a Voice Skill Masterclass, a starter pack, and built 360-degree support that helps create a good Alexa skill and a defined strategy of the next steps after the skill is built. Through our application, the users will monitor their Alexa skills data and see areas needing improvement to succeed.