1. Help Center
  2. Before you start

Best practices for conversation design

When you create your bot, it’s important to consider its purpose and persona, as well as the flow of its chats with customers. These are essential elements of conversation design. 
Elements of a chatbot supporting its purpose and persona, including avatar, brand logo, greeting message, and starting flow

In this article, we’ll share some best practices to help you enhance the bot-user experiences, addressing how to:

Establish the basics before building

Determine how and where you’ll use your bot. This helps ensure it provides interest and value for customers, supports your company’s goals, and works within your existing infrastructure.

Understand your audience

Which users are you targeting? Do you know their needs and pain points, and what you can do to address those? What is their typical user journey? How do they prefer to communicate?

Answering these questions will help you lay the groundwork for your chatbot. This information will help you tackle several of the elements we discuss in this article, such as your bot’s purpose, persona, conversation paths, and more.

Need help determining your customers’ needs? Consider using the bot to collect user data, then design and refine it accordingly.

Define your chatbot’s role

It’s crucial to establish the purpose of your chatbot early on. Is your bot’s purpose to…

  • Improve customer service?
  • Attract potential customers?
  • Act as a communication channel?
  • Enhance product sales and marketing efforts?

This role is central to your bot-building process, influencing how you establish KPIs, select integrations, determine access to a human agent, style the bot, and much more.

Imagine a clothing store that wants to launch a chatbot. They take the time to clearly define the bot’s role: improve product sales and customer support. They design a digital twin of their best salesperson that finds items in the blink of an eye, provides help with sizing and discounts, and automates customer service.

Be just as precise about the goals for your bot. The conversation it has with each customer must offer an obvious benefit. Otherwise, it’ll likely be abandoned.

Select your deployment channels

Where will your bot be used? Your company website is a starting point, but consider other channels too. Determine where users are most likely to benefit from the bot's support.

Deciding this early on will help you with the concepts we address below, ensuring your bot’s messaging and flow make sense for the places where your users will interact with the bot.

Create a chatbot persona

Your chatbot represents a chance to present your brand in an engaging, more “human” way. As such, it’s important to develop a bot persona that is aligned with your company’s identity in its appearance, tone, and behavior.

Style your bot

Ensure your bot stays true to your brand’s visual identity. Having an avatar is a great way to reinforce this. It can be as simple as a company logo, but a cartoon graphic works well too.

We recommend not using an image of a real person as it can provide the wrong idea to customers.

You can also customize the colors, fonts, and more of the Certainly Widget. Consider how you can make your chatbot stand out while still aligning with your organization's brand style guide.

Use the right language

The chatbot should reflect the tone of your brand as well. Have fun with the messaging while staying aligned with your company’s identity. Emojis are a great way to add personality!

Everyday, conversational vocabulary is best to create a more relaxed, casual environment. Keep things informative and straight to the point, and always be respectful.

Make an introduction

Human conversations typically begin with a greeting and introduction. Bot visitors expect this too, so your bot should start by saying “hi” and explaining who they are. It signals to users that the bot is not a human, even if you've designed a realistic persona like a virtual salesperson.

Personalize the conversation

Try using contextual data to customize your messages, such as your customer’s name.
Example of personalized bot message greeting a customer by name

Your bot could even switch to a different language based on the visitor’s location. This helps create a sense of familiarity and strengthens the connection with the user.

Prioritize the readability of your messaging

Pay extra attention to the readability of your bot's content to help make conversations between your chatbot and customers as effortless as possible.

Keep it simple

Short messages are easy to read and understand, and they encourage more engagement. Write simple, to-the-point sentences to keep the conversation moving. 
Side-by-side examples of bot messages, one showing what not to do with a very long text and the other recommended version that is shorter, to-the-point, and links to more info

Don’t try to pack too much in at one time! Using links to sections on your website is a good way to provide valuable info while limiting the bot message content.


Allow time for reading

Let your customers read the bot’s responses by using message delays. This adds a more natural feel to the chat flow and helps bot visitors feel like they're part of a fluid conversation.
View of Message tab when editing a Module, highlighting the feature to add a delay to bot messages

To add delays, use the timer feature within bot messages.


Include conversational markers

In a normal conversation, people use a range of cues that provide valuable context. We recommend incorporating the following in your chatbot messages:

  • Discourse markers: Guide the user by making connections between conversation parts, such as indicating what comes next or referencing how it connects to what came before.
    Examples: “Also”, “Ideally”, “Therefore”, “First”, “Otherwise”
    Excerpt from a conversation where the bot uses a discourse marker of "In that case"
  • Acknowledgments: Recognize that the bot has received the customer’s response.
    Examples: “Got it”, “Sure”, “Thanks”, “Good choice”, “Checking on that now”
    Excerpt from a conversation showing bot using an acknowledgement of "Let me see what I can find..."
  • Clear endpoints: Make sure chatbot visitors are aware when a conversation is over, and identify any next steps as needed.
    Examples: “All done! You can exit the chat 😊”, “Glad I could help! Thanks for the chat👋”
    Excerpt from conversation where bot indicates a clear endpoint by saying "Thank you for choosing us. Have a great day! You can close the chat now."

Avoid repetition

Write multiple versions of the same message for a smoother conversation, and don’t use the same version twice in a row. Consider the various scenarios that will lead to this response.
Examples of bot messages that use the following four variations to respond to the user's "Thank you": "You're welcome", "Always happy to help", "My pleasure", "Anytime"

Learn how to use the Variation feature.

Emphasize appropriately

Avoid using fully CAPITALIZED words to provide emphasis. Instead, use bold and italic text to highlight important chat details. Just remember to do so sparingly; otherwise, it loses its effect!

Ensure intuitive navigation

The ideal chatbot is helpful from the beginning, with well-planned flows and options catering to the users at every step.

Start off right

Set up clear paths that provide obvious value to the customer. Include starting flows throughout the entire conversation to guide users down specific paths, spark their interest, and point out different choices based on their expressed interests.

In the following examples from our Basic Ecommerce Bot Template, the chatbot visitor is presented with options that become more focused as the user further identifies their needs.
Bot messages that offer numbered menu options to a user to direct the conversation flow

Consider the next step

Always offer a next step, an alternative, or even the option to go back to ensure the conversation never dead ends. This could be:

  • presenting new menu items that seem relevant,
    Bot message that displays buttons to a user with other related paths to choose from
  • requesting feedback on the bot’s ability to help the customer,
    Bot message asking if the info was helpful with "Yes" and "No" buttons after providing an answer to a visitor's query
  • preparing a fallback for cases when the bot doesn’t understand,
    Bot responds to user's incorrectly spelled query with "Sorry, but I didn't understand your message. Would you mind trying to rephrase, please?"
  • or providing a follow-up alternative when no human agent is available upon request.
    Bot message offering an email address and phone number for customer support when an agent is not available

Be proactive

Have the bot take the initiative in the conversation with proactive support.

  • Pop-up messages can help you attract users to a conversation or help them when you think they’ll most need it.
    Proactive messages shown above chatbot widget on store's website
  • Use questions with Suggested Replies to obtain more information from the customer, qualify it better, and help users stay on the right path.
    Bot message accompanied by three selectable Suggested Replies for the visitor to choose from
  • Consider adding a persistent menu to get your visitors back on track from any point in the conversation.
    Three vertical dots menu shown to the left of the chatbot widget's text entry field used to access the persistent menu

Evaluate and improve performance

As a bot-builder, you’ve worked hard to create a bot that provides value! Over time, you’ll want to monitor the bot’s performance and conduct regular maintenance to maximize your return.

Establish an evaluation procedure

Reviewing your bot's conversations with visitors will help you locate potential areas of improvement. We recommend using the Message Insights Report to identify chats where the bot doesn’t connect with your users, understand why, and take action.

Make comparisons

Compare your bot to both itself and competitors.

  • Return to the role and goals you established for your chatbot at the beginning of the project. Has anything gotten lost over time?
  • Gain a better understanding of how your bot is performing with our Traffic Report. What changes can be made to better serve your customers?
  • Research what others in your field are doing. How can your chatbot be refined to outperform the competition?

Remain vigilant, as bots are still a relatively new technology.

Optimize and improve your chatbot

Use the information you’ve gathered from your evaluations and comparisons to keep improving your chatbot after every conversation. Constantly assessing the bot’s performance will enable you to optimize step by step, add new features, or enhance those you’ve already built. 

Remember that nothing is perfect, and everything is subject to improvement!

Need help interpreting your data? Have ideas for optimization yet are unsure about the implementation? Just reach out to our Customer Success team – we’re always happy to help.