Published: April 14, 2023
Key takeaways
Alexa skills that are meant to be used repeatedly and retain customers over a long period of time should be designed to fit the needs of those customers throughout their journey.
Follow the best practices in this article to make sure your skill feels fresh to customers their first visit, or their 1,000th.
Need quick advice?
View the Checklist for designing a skill for retention.
In this article:
What is retention?
It’s likely you want people to use it more than once. Ideally, customers will love your skill enough to continue using it over a long period of time, a process called retention. A measurement of retention refers to the percentage of customers who continue using your skill over time, and it’s one of the best indicators of skill quality: Customers only return to skills repeatedly over time (retention) if they enjoy and derive value from them.
Successful skills don’t necessarily implement all of the practices discussed here, but they combine several to maximize the chances their customers will remember and continue to enjoy their experience long-term.
Checklists for designing a skill for retention
Consider your value proposition:
▢ Your skill should offer content or utility that is valuable, useful, or entertaining to customers
▢ The content should be abundant, fresh, and/or dynamic enough to sustain repeated use
▢ Don’t rely on unsolicited messaging in the skill (random suggestions or promotion); any upsells for ISP content should follow best practices
Reduce friction:
▢ Choose & test an invocation name that is easy to remember, easy to say, and that Alexa recognizes with a high degree of accuracy; to learn more, read Choose the Invocation Name for a Custom Skill
▢ Take as few turns as possible required for the customer to reach the core content of your skill
▢ Prevent errors; handle them gracefully when they do arise
▢ Check that there are no dead ends in the possible paths of dialog
▢ Be brief: Most responses should be short enough to be read by a human in one or two breaths
▢ Make the skill accessible to all
▢ Follow conversation design best practices, especially principles of natural speech
Be dynamic:
▢ Greet the customer with a dynamic message
▢ Change the design and/or messaging in your skill based on seasonality, customer milestones, or other events
▢ Vary dialog the customer will hear often
▢ Offer contextually relevant suggestions; don’t offer options that aren’t relevant
Experiment & respond to feedback:
▢ Experiment often with new features or alternative ways of completing a task
▢ Thoroughly test your skill before you submit for certification with a large variety of people
▢ Monitor customer reviews in the Alexa skills store
▢ Monitor your Alexa Skill Analytics dashboard
▢ Check your Skill Quality Score
For customers to remember and continue using your skill over time, it will need to offer a reason to return. Your skill’s purpose, content, and utility should be compatible with frequent use. Customers can’t come back to your skill daily, for example, if there isn’t something new to do every day.
Skills that retain customers at a high rate tend to do one or more of the following …
They contain actions tied to frequent habits. When customers make a skill a part of their daily habits, they’re most likely to return daily with the right incentives. Examples include:
- A meditation skill helps customers build healthy habits with a daily meditation each morning, and a calming ambient sound session for sleep at night.
- A child-directed skill helps children brush their teeth for the right amount of time each morning and night.
- A nutrition skill helps customers log, track, and reach their daily water intake goals.
- A health & wellness skill helps customers start a warm-up and cool-down routine before and after their regular home workouts.
- A trivia game helps customers learn about houseplant care while they water their plants
They appeal to a sense of accomplishment. When customers believe they’re participating in something greater than themselves, they’ll be motivated to return to your skill to fulfill those goals. Examples include:
- A daily sports trivia skill allows fans to earn points for their favorite team. Fans are informed who “won” each week.
- A trivia skill with daily challenges tracks and surfaces a score leader board to all players.
- An entertainment skill asks customers for their favorites across different topics, then reveals how others answered.
They enable personalization & ownership. Alexa is a personal assistant. Your skill should be as contextually aware as possible and become more relevant to customers over time. Allow customers to personalize their skill experience directly and indirectly, and leave indicators the skill is “theirs.” Examples include:
- A sports-themed skill welcomes customers with an update on their favorite team, or the team the customer last asked about .
- A fitness skill collects customer preferences and goal to mix a customized workout.
- A recipe skill remembers a customer is vegetarian for future reference, and doesn’t surface search results that contain meat again, unless asked.
They create anticipation, curiosity, and surprise. Customers return to skills that build anticipation for the next interaction, and they don’t want to miss limited-time content. People are also curious and will keep interacting to see (or hear) what the skill will do next. A surprised and delighted customer will return to a skill to continue to be surprised.
- A skill surfaces a different greeting each day tailored to the customer’s status, such as telling them about their upcoming fitness class, or who their favorite team is playing next.
- A customer is welcomed into a game skill by being rewarded for a recent achievement or milestone with a round of bonus questions.
- A choose-your-own-adventure story ends for the day on a cliffhanger. A listener will return tomorrow to learn what happens next.
They enable creativity and encourage feedback. Skills that elicit the customer’s creative input and letting them experience the outcome of their decisions is more engaging than those skills that don’t adapt to the customer’s inputs. Examples include:
- A skill surfaces suggestions for recipes according to the customer’s answers to a few personality and dietary questions
- An interactive story offers a number of choices to influence its outcome, for better or worse.
- An ambient sound skill asks customers about their mood and preferences to mix a custom soundtrack.
Example: You can experience a simple simulator skill that is built to retain customers by playing the MyFarm skill by Amazon.
Checklist to consider your value proposition:
▢ Your skill should offer content or utility that is valuable, useful, or entertaining to customers
▢ The content should be abundant, fresh, and/or dynamic enough to sustain repeated use
▢ Don’t rely on unsolicited messaging in the skill (random suggestions or promotion); any upsells for ISP content should follow best practices
To learn more about the kinds of skills customers love, read Design Principles: Should you use Alexa?
After you’ve considered how your skill’s concept and content will offer value to your customers over time, you should consider how levers outside the skill session could drive them to return, and return more often. There are APIs and features you can implement that can help boost engagement and retention, either directly or indirectly.
Some levers you might consider include …
Once you’ve decided on a skill concept and content, and considered levers you might use to maximize your regular visitors, it’s time to design your skill experience itself for those repeat customers.
There are a few hallmarks a skill should have when it’s meant to retain customers who use it frequently over a long period of time. These hallmarks all help keep skills interesting, engaging, and adaptive to the needs of the customer.
Be friction-less
People won’t return to a skill that’s too difficult, unreliable, or tedious to use. Friction deters customers from coming back to your skill. It might drive them away immediately in the case of repeated errors or too many steps, but it might also drive them away after a period of time through repeated minor annoyances.
There are a few kinds of friction a customer might encounter:
- Invocation issues: Customers have trouble invoking your skill. When you choose your skill invocation name, make sure you also understand how customers invoke skills.
- Errors & dead ends: Customers reach repeated errors, even errors that will end the skill session. Customers may also reach messages that may not indicate what they can say or do next. The skill doesn’t ask a question or open the mic to elicit a response. Learn more about how to prevent and handle errors.
- Confusing prompts: The customer doesn’t understand from your skill’s responses what your skill does or what they can say next to accomplish a task. Learn more about writing natural speech for Alexa.
- Verbose dialog: Customers become annoyed if a TTS response is longer than what would take a human to say in one breath (maximum two). Learn more about writing brief speech for Alexa.
- Repetitive dialog: If Alexa says “Ok.” or “Sorry” before every message, or says the same thing each time the customer visits, the skill will sound increasingly impersonal and robotic over time. Vary the dialog customers will hear repeatedly. Randomize greetings, salutations, and short phrases.
- Trustbusters: The skill doesn’t live up to customers’ expectations of what it should do or what they should be able to say to the skill. The skill doesn’t protect sensitive information or take preventative measures to avoid errors. Learn more about designing a trustworthy skill.
- Inaccessibility: The skill isn’t accessible to all customers, including those with a range of disabilities. It relies on touching the device or the customer’s ability to see a screen to function. Legibility of the on-screen display is poor. Learn more about designing skills accessible to all.
Checklist to reduce friction:
▢ Choose & test an invocation name that is easy to remember, easy to say, and that Alexa recognizes with a high degree of accuracy. To learn more, read Choose the Invocation Name for a Custom Skill
▢ Take as few turns as possible required for the customer to reach the core content of your skill
▢ Prevent errors; handle them gracefully when they do arise
▢ Check that there are no dead ends in the possible paths of dialog
▢ Be brief: Most responses should be short enough to be read by a human in one or two breaths
▢ Make the skill accessible to all
▢ Follow conversation design best practices, especially principles of natural speech
Offer timely & dynamic content
Diverse, engaging, and frequently updated content is another critical element to ensuring long-term customer delight. People shouldn’t run out of content or things to do in your skill within a couple sessions, and the content it offers should be timely. If it seems to the customer the skill is offering them content they’ve already heard, or content that’s out of season, they might not return.
For example, when a customer asks “Alexa, where’s Santa?” in July, the NORAD tracks Santa skill will tell them Santa is on vacation. If they are using an Echo Show, it displays a scene with Santa either camping in the woods, or surfing at the beach. However, if they ask the same question on Christmas Eve, they’ll be shown a map and told where Santa is currently delivering presents.
You should also make sure your skill supports multimodal devices with touch interactions, as well as Fire TV devices with remote control interactions. When it’s appropriate, your skill could also offer an experience for those with On-the-Go devices and/or Echo Auto.
Personalize responses & experiences
Demonstrate to customers that the skill has considered previous requests in a recommendation, and don’t offer something that isn’t relevant to them. There are a few ways you can make your skill more personalized and relevant to each customer:
- Important information comes first: Give customers the information they’re most likely to need most often early in their experience; other important information should be easy to request.
- Dialog is contextually relevant: Don’t offer options or features that aren’t relevant or available to the customer.
- Offer rewards, recognize milestones & achievements: Recognize and reward repeat visitors with special content, greetings, or other rewards.
- Reference past experience: Remember preferences or allow customers to manipulate settings directly to influence the kind of content the skill will surface.
In what ways has the (hypothetical) skill example below personalized the experience for this particular customer?
Customer: Alexa, open Plant Pal.
Brian: Welcome back to Plant Pal.
<audio src="soundbank://soundlibrary/ui/gameshow/amzn_ui_sfx_gameshow_tally_positive_01"/>
Brian: Congratulations. It’s your 10th day tracking your plants with Plant Pal. You’re on your way to becoming a real green thumb!
<audio src="soundbank://soundlibrary/ui/gameshow/amzn_ui_sfx_gameshow_positive_response_02"/>
Brian: Looks like six of your plants might need urgent watering. Do you want to get started?
Customer: No not yet.
Brian: Ok. I have a new trivia question about your plants. I bet it’ll knock your stalks off.
<audio src="soundbank://soundlibrary/musical/amzn_sfx_drum_comedy_02"/>
Brian: Would you like to play?
Customer: Sure.
Brian: You got it. Today’s plant will be familiar to you. It’s the Alocasia. But how many species are there? Is the answer A. 1,300 species, B. 90 species, or C. 350 species.
Customer: I think it’s 350
<audio src="soundbank://soundlibrary/ui/gameshow/amzn_ui_sfx_gameshow_negative_response_01"/>
Brian: Not quite. The answer is A. There are about 90 species of Alocasia. Not everyone agrees on the classification of a few. That’s a score of 8 out of 10 so far. Not bad.
Brian: A few of your plants may need your attention. Are you ready to get started?
Encourage discovery & follow-up
Skills that have lots of features or capabilities can be difficult for a customer to understand, especially at first. You’ll want to create dialogs that make it clear what your skill can do. Keep in mind that many of your customer’s won’t have a screen to look at, so you’ll need to include everything they need to know in the skill’s verbal responses when a customer is using a device without a screen. So, what kinds of dialogs can you design into your skill to help customers discover your skill’s whole potential?
Greet the customer dynamically: You’ll want to introduce your skill differently depending on whether your customer has visited before, or never. What if they’ve visited hundreds of times? Examples might include:
- A banking skill would tell a first-time visitor that after they link their account, they can check their balance or credit score, or pay their bill. After they link their account and return, the skill tells them their credit card balance and due date. How would you greet a customer who already paid their bill that month?
- A trivia skill would tell a first-time visitor about the rules and scoring of the game, while a player who is returning for a tenth day in a row is rewarded with a special greeting and a few extra trivia questions to play that day. How would you greet a player who returns to your skill but hasn’t played in a long time?
- A sports-themed skill greets a first-time visitor by asking their favorite team. On subsequent visits, the skill will greet the customer by telling them either the score of a game that team played within the last 24 hours, or the date and time of their next game and the opposing team. How would you keep the content fresh for them in the off-season?
In what ways does the following skill example help this customer discover what the skill can do?
Customer: Alexa, open Sports Wire.
Matthew: Welcome back to Sports Wire. The Seahawks swiftly defeated the Cowboys last night 37 to 9. Want to hear the highlights?
Customer: When do they play next?
Matthew: The Seahawks play at the Ravens next Sunday, January 23, at 3 pm. Should I remind you a few minutes before kickoff?
Customer: No thanks.
Matthew: Ok. I can help you find another game, listen to live sports radio, and more. What would you like?
Customer: Play the radio
Matthew: Sure. Here’s Seattle Sports Talk 92.5. By the way, you can ask Sports Wire to change the station any time.
<Radio plays>
Offer contextually relevant suggestions: Use what you know about the customer through the information available to you and their usage history with your skill to make their experience easier or more relevant. Examples could include:
- A skill that helps customers book fitness classes suggests an upcoming class the same time and day they took the same class last week.
- A skill that helps customers search for recipes lists gluten-free options first when the customer asks for a recipe, since they have asked for and rated gluten-free recipes in past sessions.
- A skill that helps customers look up flight information offers updated information about a flight the customer searched for in a previous session that has not yet departed.
- A game skill that simulates raising a plant suggests actions the player can take to fulfill the plant’s needs, such as watering a thirsty plant, or raising the blinds for sunlight.
In what ways does the below skill example offer relevant and contextually appropriate options to the customer?
Customer: Alexa, ask Lunch Buddy to order a sandwich.
Alexa: Welcome back to Lunch Buddy. Last time you had a Veggie wrap with extra olives and double provolone cheese. Would you like to order it again?
Customer: No, I want something else.
Alexa: Ok. I found a few more popular vegetarian options, including the chickpea salad sandwich, made with vegan mayo. Want to try it?
Customer: How about the muffaleta?
Alexa: The muffaletta sandwich comes with classic cold cuts, provolone, and sesame seed bun. Just so you know, it’s not vegetarian. Want to customize it?
Customer: Give me that chickpea salad one.
Alexa: You got it. One chickpea salad sandwich. Want to customize anything?
Customer: Nope.
Alexa: Ok. Ready to check out?
To learn more about creating contextual conversations, read Design Principles: Be Contextual.
Be proactively helpful: Offer additional explanation when the customer makes the same mistake more than once, give them hints or examples of things they might say, and offer contextually relevant help if the customer asks for it.
- A skill that helps customers order food might offer some additional explanation that they can only add up to five extra shots of espresso in their coffee, not the 25 the customer just requested.
- A trivia skill might help a customer avoid an unintended penalty in their score by reminding them the ways they can answer when they reach their first error or “wrong” answer.
- A skill that helps customers track and increase their water intake responds to a customer’s request to “log a glass of water” by letting them know the various ways that they can describe the volume of water.
In what ways has the skill in the following example tried to proactively help the customer?
Customer: Alexa, ask Fit Friend to log a glass of water.
Kendra: How much water did you have?
Customer: A really big glass?
Kendra: I can log water using ounces, pints, gallons, cups, liters, and more. How much water did you have?
Customer: A couple pint glasses.
Kendra: Two pints. Logged. You’re halfway to your daily goal. Do you want to hear more about your log?
Customer: I actually had three pints.
Kendra: Three more pints. Logged. You’re almost to your dail––
Customer: Alexa, that’s not what I meant.
Kendra: Hmm. I’m not sure about that. Want to hear more from your water log?
Customer: Help
Kendra: You can ask me to log water, or I can tell you what's in your log. if you want to make a change, you can remove water or clear your log for the day. What would you like to do?
Customer: remove two pints of water.
Kendra:Two pints. Removed. Anything else to log?
Customer: No that’s it.
Kendra: Thanks for being a fit friend. Should I remind you to log your water at 10 am again tomorrow?
Customer: No thanks.
Kendra: No worries. Have a hydrated day.
To learn more about offering help and handling errors, read Design Patterns: Handling Errors.
Follow up: If the customer successfully completes a task, where possible, offer an option to do something else that is most related to their previous task. Don’t offer tasks the customer wouldn’t need to complete, such as the same task they just finished.
- A skill that helps customers track flights could follow up some flight information with an offer to send the customer a notification if the departure time or gate changes.
- A skill that helps customers book salon appointments might follow up a successful booking with an offer to remind the customer of their appointment an hour before.
- A skill that helps customers order food might follow up a description of the daily special with a prompt asking “Want to add it to your cart?”
In what ways has the skill in the following example followed up with the customer to help them continue using the skill?
Customer: Alexa, open MyFlight Finder.
Alexa: Welcome back. Your Alaska flight 1601 to Melbourne Florida, leaving today at 12 pm from Seattle, is delayed. It doesn’t have a new departure time yet. Want me to notify you of any more changes to your flight?
Customer: No thanks.
Alexa: No problem. Want to check another trip?
Customer: What other flights to Melbourne are there today?
Alexa: I found flights to Melbourne from Seattle today departing at 3 p.m., or 7 p.m. Which do you want?
Customer: That’s it I’m done.
Alexa: Ok. Happy travels.
Kendra: No worries. Have a hydrated day.
Checklist for dynamic skills:
▢ Greet the customer with a dynamic message
▢ Change the design and/or messaging in your skill based on seasonality, customer milestones, or other events
▢ Vary dialog the customer will hear often
▢ Offer contextually relevant suggestions; don’t offer options that aren’t relevant
Test your skill before you submit it for certification to ensure at least a small sample of customers can move as expected through the possible dialog paths as intended. Gather their feedback and tweak your design as needed. After the skill goes live, monitor Alexa Skills Store reviews, and use Alexa Skill Analytics to ensure a larger number of customers are still using your skill without issue. If you make any changes in response to skill reviews, make sure you update your description in the Alexa Skills Store so others know it’s been updated.
You can also use a number of tools in the Amazon Developer Portal to see how customers are using (or not using) your skill:
- Analyze interaction paths with the Interaction Path Analysis tool available on the Analytics tab to learn how customers are using your skill. It can help you identify interactions where customers become blocked or commonly exit the skill. You can use insights gained from interaction path analysis to make your flow more natural, fix errors, and address unmet customer needs.
- Check your Skill Quality Score and follow the suggestions in the Skill Quality Coach. Learn more about the Skill Quality Coach.
- Use the Intent History page to learn when your skill surfaced errors and noted other intents like “stop.” It may help you find a problem that causes customers to leave a skill mid-session.
You’ll also want to experiment when you’re unsure how to solve a problem, want to introduce a new feature, or are unsure which solution among many might best solve a problem. For example, if your skill will offer monetization, you can complete an A/B test to see which in-skill upsell message performs the best. Experiment frequently once you’ve launched your skill to fine-tune your the experience and make your dialogs their most effective.
To learn more about getting feedback from your customers, read Design Patterns: Gathering Feedback.
Checklist for experimenting & responding to feedback:
▢ Experiment often with new features or alternative ways of completing a task
▢ Thoroughly test your skill before you submit for certification with a large variety of people
▢ Monitor customer reviews in the Alexa skills store
▢ Monitor your Alexa Skill Analytics dashboard
▢ Check your Skill Quality Score
Tip: Use different voices & sound in your skill
Some skill examples on this page used voices that weren’t Alexa’s, as well as sound effects. Another way to prevent a skill from becoming dull to customers over time is to use another voice, or assign different “roles” spoken at various points in your skill to different voices. While you shouldn’t do this for its own sake, using another voice that better matches a personality you’re trying to convey, or using multiple voices to help the customer distinguish between types of content or functions in your skill can also help improve retention and keep your dialog sounding fresh. (Customers hear a lot from Alexa, after all). If available, high-quality, pre-recorded audio can bring maximum impact.
To learn more about using different voices in your Alexa skill, read Improve your audio experience