Best Practices

A well-configured chatbot can be a powerful asset. Following these best practices will help you create a more effective, accurate, and user-friendly AI assistant.

Curate Your Knowledge Base

The quality of your chatbot's answers depends entirely on the quality of your source material. Garbage in, garbage out.

  • Be Specific and Clear: Use documents with clear headings, simple language, and a logical structure. Avoid long, rambling paragraphs or marketing fluff.
  • Start Small, Then Expand: Begin with your most important content, like an FAQ page or key product documents. Test the results, then gradually add more sources.
  • Use Q&A for Precision: For critical information like your return policy or specific pricing details, use the manual Q&A source type. This gives you exact control over the chatbot's answer.
  • Keep It Updated: If your product or policies change, remember to update or re-sync the relevant data sources to ensure your chatbot provides current information.

Design a Great User Experience

The chatbot's interface is the first thing your users interact with. A thoughtful setup can significantly improve engagement.

  • Write a Welcoming First Message: Your welcome message should greet the user and briefly explain what the chatbot can help with. For example: "Hi there! I'm the Acme support bot. Ask me anything about our products, pricing, or policies."
  • Use Smart Suggested Questions: Populate the suggested questions with your top 3-4 most common user queries. You can find these on your Dashboard's "Top Questions" panel.
  • Match Your Brand: Use the Customization page to adjust the chatbot's color and avatar to match your website's branding for a seamless, professional look.

Monitor and Improve

Launching your chatbot is just the beginning. Regularly reviewing its performance is key to long-term success.

  • Review Conversations: Spend time each week looking through the Conversations page. It's the best way to understand what your users are asking and how the chatbot is performing.
  • Analyze "Top Questions": The Dashboard shows you the most frequent topics users ask about. If a top question isn't being answered well, it's a sign that you need to add or improve a data source covering that topic.
  • Pay Attention to Sentiment: A drop in the "Positive Sentiment" score is a clear indicator that users aren't finding the answers helpful. Investigate recent conversations to understand why.

The goal is a continuous feedback loop: Monitor conversations, identify knowledge gaps, add new data sources, and repeat.

Test Like a User

Before and after you make changes, always interact with your chatbot from the perspective of a new user.

  • Ask Ambiguous Questions: See how the chatbot handles queries that aren't perfectly phrased.
  • Try to Break It: Ask questions you know are outside the scope of its knowledge. A good response is "I don't have enough information to answer that."
  • Use the Live Preview: The live preview on the Customization page is your sandbox. Use it extensively whenever you add new knowledge or change settings.