Building a knowledge base on tawk.to: Deciding what to write first based on customer testimonials and search data.
When setting up a self-help center, the most difficult part is often determining the first batch of articles, not choosing templates. Customer service dialogues, common search terms, and existing instructions can all be used to determine priority.

Starting with recent real-world questions, we collected direct feedback from both buyers and frontline customer service representatives. We combined various question formats into a single task, such as "Can't find the download link" and "Where do I get the file after payment?" We retained commonly used customer terms as titles or search terms, and the body text directly answered this task.
Review frequently searched terms and readership in the knowledge base, then compare them with chat and email records. Topics with high search volume but low resolution may already have articles but lack clear steps; high-frequency topics with no pages at all are the new candidates. Internal training materials, installation guides, and product manuals can be reused initially, while removing customer data unsuitable for public release.
For multi-step operations, clearly describe the prerequisites, actions, and the interface status after success. Have a colleague unfamiliar with the system follow the documentation once, noting which step they are on, so they can still contact customer service. After publishing, review the number of related work orders and search feedback, only adding truly missing information each time.



