Establish and Maintain Trust

Customer engagement is built on a foundation of trust. By maintaining a trusted relationship with your customers, you can build a loyal customer base for your skills and devices, which helps to grow Alexa's credibility for customers as a virtuous cycle. Any good or bad interaction can lead to the breakdown or buildup of trust, and bad interactions can lead to lower engagement, bad publicity, and low customer satisfaction. The following sections outline the type of interactions that can build up or break down customer trust.

Interactions that break trust

Offensive content

It is inevitable that some customers ask offensive things that you need to handle appropriately in your skill. While it's impossible to control what a customer says, it is likely that some customers look for or find sensitive or offensive topics. While you can make a skill with offensive jokes or other material, you might end up alienating your customer base. What is defined as offensive or sensitive changes by country, region, and person. Be cautious when translating or localizing content from one locale to another.

Sensitive topics
A voice experience shouldn't violate a customer's expectations of a skill's content, and the customer should explicitly ask to engage with such content. Skills not rated as mature, for example, should avoid telling a joke with sexual innuendo even if the language contains no obscene words. If you have a skill that contains mature content or profanity, you should still avoid surfacing that content immediately when the customer invokes the skill, as he or she might have invoked it in error.

Profanity and derogatory terms
When a customer utterance results in an Alexa response that contains profane or derogatory terms, you should redirect the customer with a re-prompt or have Alexa answer them politely with, "I'd rather not answer that.”

Sexual or medical content
Alexa isn’t a doctor and shouldn’t advise any treatments or diagnoses. You need to take these kinds of requests seriously; Alexa should never respond with a humorous answer. Instead, provide only factual, non-judgmental responses, even to customer requests that might be perceived as sexual or objectifying to customers’ body image, gender identity, sexual preferences, or disabilities.

Politics and religion
When customers raise topics that involve an ideological dispute, geographical or cultural dispute, or possible offense, you need to handle these topics on a case-by-case basis with sensitivity.

Artistic sources
Be careful with referencing creative works such as music, books, songs, or movies that contain profanity in the name or title. The profanity could also be contained in the artist’s name, quotes from their work, or articles about them and their work.

Unsolicited content

Despite your best intentions, when you surface information that is unrelated to a customer's primary task or mention products or brand names, customers may perceive this as promotional content that puts your business needs over their customer needs. This can lead to a breakdown of trust and reliability.

Unwanted advertisements
While you can promote items for sale in your skill or other related skills, be careful not to advertise items that don’t fit within the context of your skill. Avoid unwanted purchasing suggestions, such as providing the same item for purchase over and over when the customer has declined the item. For more information about selling products through Alexa, see How to make a purchase suggestion. Learn more about selling products through Alexa.

Unintentional promotions and secondary tasks
You might be inclined to enhance the customer's primary task through things like proactive messaging, feature awareness, or quirky quips. While your intention is to improve the overall experience, customers might misinterpret these messages as pushy advertising or unwanted promotions. Give customers a reason why you’re suggesting the task or item and how it aligns with their goals for that skill.

Advertising or promotions child-directed skills
Child-directed skills have their own rules and regulations, such as the FTC's COPPA Rule, to protect children from unwanted content such as advertisements. For more information about the Alexa policy for child-directed skills, see Policy Testing.

Unmet expectations

Alexa should deliver on any task or item that she promises to deliver.

Not completing desired tasks
Alexa should prioritize fulfilling customer requests over delivering messages unrelated to customer needs. Not every action requires Alexa to confirm it. For example, when a customer asks to stop or exit a skill, the skill should end the session without confirmation. Alexa shouldn't ask for permission unless there's a compelling reason, such as losing significant progress in a game.

Requiring customers to link an account unnecessarily
Skills should deliver as much content as possible to a customer without requiring account linking. When customers encounter the added friction of account linking to do a simple task, they may abandon the skill and do that task elsewhere. Customers might distrust skills that won’t let them perform simple tasks without having to provide additional data through account linking. Use the additional data provided by account linking to delight customers and reduce friction, while still meeting expectations and providing value to customers who haven't linked their accounts yet.

Technical or feature limitations

Technology or feature sets might not always meet customer expectations. When you design your skill, try to anticipate features that customers want and look at customer data to determine if you need a help prompt or prioritization mechanism.

Technological limitations
A customer might ask, “Alexa, play the artist Japanese House.” Instead, Alexa continues to play the music genre “Japanese house.” Should you encounter feedback for items like this, consider applying quick short-term changes to address issues that have major customer impact.

Settings and privacy
Avoid announcing or displaying private or sensitive information or content without the customer's consent. This might include cases such as Alexa repeating back passwords or security PINs, or displaying bank account information on a device screen.

Skills that rely on location for some functions can ask the customer for permission to access that data. If the customer denies permission, or hasn't provided the data, the skill should offer alternate functionality or collect the data in some other way, such as asking for a zip code to locate a business nearby.

Frequent errors
Errors and latency might affect a customer's trust. Design your voice experience to handle situations that might result in errors; otherwise, you might accidentally design interactions that exacerbate the problem.

Guidelines for building trust

Thoughtful prompt writing is your path to building trust:

  • Be honest but remain useful
  • Provide clear information about errors
  • Confirm high-risk tasks
  • Maintain consistency and context

Use confirmations in high-risk situations

Explicit confirmations

These confirmations require customers’ verbal approval to continue. It adds an additional step in the conversation but protects against high-consequence failures such as monetary impact, data or game progress lost. The explicit confirmation acts as a safety net.

Do

User: “Alexa, ask Banker to transfer four hundred dollars from my checking to my savings account.”

Alexa: “That was a transfer of four hundred dollars from checking to savings account, correct?”

Don't

“Alexa, ask Banker to transfer four hundred dollars from my checking to my savings account.”

“Okay.”

Implicit confirmations

These confirmations re-state what customers said, but don’t ask them to confirm it. The customer can take action to stop the interaction and correct the mistake. The likelihood of a mistake is low, and the consequences of a failure are minor.

Do

User: “Alexa, ask Trip Planner to start my vacation on June ninth.”

Alexa: “Okay. You're leaving June ninth. What day are you returning?”

Don't

User: “Alexa, ask Trip Planner to start my vacation on June ninth.”

Alexa: “You said June ninth, right?”

Contextual confirmations

These conformations let customers know that the skill understood them and that they are moving to the next step in the skill workflow. The confirmation doesn't explicitly re-state the utterance, but implies that Alexa understood correctly based on the next response it delivers. Use these for faster-paced conversations with low probability and minor consequences for errors.

Do

Alexa: “Which category would you like? People, places or animals?”

User: “Animals.”

Alexa: “For your first question, what animal is the only marsupial in America?”

Don't

Alexa: “Which category would you like? People, places or animals?”

User: “Animals.”

Alexa: “You selected the category Animals. For your first question, what animal is the only marsupial in America?”

Use informal, conversational language

Use common, familiar language that any customer can understand and avoid technical jargon.

Do

User: “Alexa, ask Animal Sounds to play Safari Radio.”

Alexa: That radio station isn't live right now.

Don't

User: “Alexa, ask Animal Sounds to play Safari Radio.”

Alexa: Sorry, I can't play the Safari Radio station because I couldn't establish a connection with the Animal Sounds server.

Don't make false promises

Avoid setting false expectations about existing or future capabilities. ‘Yet' implies that a feature might be available at a future date when in fact it might never be supported. When a feature is limited, tbe honest about it, and inform the customer of what they can do.

Do

User: “Alexa, ask MyHealthPro for my average calorie intake this month.”

Alexa: “I can't track calories by month, but I can tell you about your calorie intake yesterday, today, or record a food to your food journal. Which would you like?”

Don't

User: “Alexa, ask MyHealthPro for my average calorie intake this month.”

Alexa: “Sorry. I can't do that yet, but I'm always learning.”

Building trustworthy visual designs

As with voice design, consistent and reliable visual design that a customer can interact with reliably can build trust and engagement.

Avoid these visual design traps

  • Don't display a task or screen that is unrelated to the current customer interaction.
  • Don't display personal information such as private messages or events without asking for permission or providing useful functionality.
  • Don't include information on screen that isn't directly related to completing a task, as the information could be perceived as an upsell.
  • Don't try to sell something to a customer in error messages with a visual response.

Visual display guidelines

  • Display information on prices, alarms, timers, and calls. Position the information in a way that is obvious to the customer.
  • Generate expectations that Alexa can live up to. For example, if the visual element looks like you can tap on it, allow the customer to tap on it.
  • Honor the customer's method for interacting with Alexa. If the customer starts an interaction by voice, respond with voice. If by touch, respond visually.
  • Consider how the visual elements might augment or complement the voice response, as opposed to restating it verbatim. Can the visual elements provide additional details that are easier to see than hear, or are hard to remember? Be aware that customers might not see the visual response, so include critical information in the voice response.


Back to top