Conversation Design Best Practices

Sofie Meyer Updated by Sofie Meyer

Conversation design best practices

Design an effective chat service by using best practices from conversation design.

⏮️ Before onboarding

  1. Is the value proposition of your service clear to the user?
  2. Are any costs or incentives or immediate benefits (like compensation for time) clearly communicated?
  3. Is the chatbot’s primary purpose and what it can and cannot do clearly communicated?
  4. Have you considered leveraging existing trusted community networks or leaders to introduce your service and build initial credibility?

🌍 Language

  1. Does the chatbot use a language that is appropriate for your target user?
  2. Can the user select their preferred language?
    1. Does the bot use language that is truly localised, including dialects, emojis and possible colloquialisms (not just a literal translation)?
    2. Is the bot humble if translation and localisation is imperfect, and can users connect to a human operator if needed?
  3. Can the user switch languages if they want?
  4. Are key points explained in simple terms?
  5. Is the text simple, concise, and easily comprehended by users who may not be fluent in the language the bot is using?

🗯 Tone of voice

  1. How does your chatbot use language? (gentle, professional, funny?)
  2. Is the tone of voice consistent throughout?
  3. Does the chatbot use emojis? (Be mindful of potential data usage concerns and prioritise simple text if data is a concern) 🚀🌟💡🎈
    1. If using emojis, are they clear to the user?
  4. Does the tone of voice shift appropriately with changing between different topics (e.g., from friendly greeting to sensitive data collection)?

😊 Persona

  1. Does your chatbot need a persona?
  2. If yes, have you defined it clearly? (for guidance, please see this playbook):
    1. Target audience (demographics)
    2. Type of support (instructions, reminders, safe space?)
    3. Role (guide, assistant, peer, expert?)
    4. Personality (e.g. fun, serious, compassionate, silly)
    5. Values (action, compassion, clarity?)
    6. No-gos (jargon, formality?)
    7. Ideal emotional impact (reassured, informed, cared for?)
    8. Is there a real-world or fictional character that resembles your bot?
    9. Gender, backstory, name, age (if needed)
    10. Avatar (if needed)
  3. Have you co-created and validated that the persona is relatable and trustworthy for your users with members of the target community?

👋 Onboarding

  1. Have you clearly explained what your bot offers/can and can’t do?
  2. Does your bot explain your data privacy policy (how long it’s stored for, how it’s used) in simple, clear terms?
  3. Are you fully transparent about who your organisation is and why you are contacting the user in the first message?
  4. Do you gather informed and active consent to send WhatsApp messages to the user? (either during onboarding onto the chatbot, or elsewhere)
  5. Do you inform the user about how to change their notification preferences?
  6. Does your bot offer something of value, quickly?

👍 Interactions

  1. Is your chatbot concise? (ideally no more than 3 sentences per message)
  2. Does each step in the journey have a choice or action for the user to do?
  3. Do you know if data is cheap or expensive for your users, and have you taken this into account in designing interactions?
    1. If data is cheap, is the content format varied? (e.g. videos, Q&A, infographics, audio, images, quizzes)
    2. If data is expensive, do you prioritize text-only responses and offer richer content (e.g., videos, images) only as an opt-in download?
  4. Can users stop the conversation if they want to?
  5. Do you create natural pause points to allow for the user not to become overwhelmed?
  6. Is navigation of the bot easy and intuitive for users?
  7. Have you avoided complex UI elements (like extensive lists or highly formatted messages) that may not display well or be navigable on feature phones or older smartphones?
  8. Have you designed for interruptions, allowing users to easily pick up a conversation where they left off without losing context?
  9. Does your chatbot take advantage of the opportunity to have two-way communication for service delivery (e.g., reporting issues, applying for services, etc) and is not just an infobot, FAQ bot, or survey tool? Does the journey feel personalised to the user?
  10. Can the user easily screenshotted or otherwise remember key information for later/offline use?

🚪 Exits and errors

  1. What happens if a user types something the bot doesn’t understand?
  2. If a user gets stuck, how can they get unstuck and return to start, main menu or exit and later resume easily?
  3. When the user has completed a journey, do they receive an exit message?
  4. After a journey is complete, does the bot specify how the user can re-engage?
  5. Is there always a clear and accessible path for users to connect with a human operator if the bot cannot help or if they prefer human interaction?

🆘 Feedback and help

  1. Does your bot offer chances to share feedback?
  2. Does your bot offer users to connect to a human operator instead?
  3. Does your bot offer an alternative channel if the experience becomes frustrating?

🧪 Testing

  1. Have you gone through the whole flow and read it aloud?
  2. Have you role-played your chatbot with another person?
  3. Have you gotten feedback from at least 10 members of your target community in real-world conditions?
    1. Have you made improvements to your bot based on this feedback?
  4. Have you tested the chatbot on a different phones, including older Android models and feature phones that may be common in your target regions?

Was this article helpful?

Launching at Scale? Here’s a checklist

Contact