In this article
- The knowledge base reality check
- Writing articles that work
- Use the exact words your customers use
- Answer the question in the first paragraph
- Use numbered steps for processes
- Screenshots for anything visual
- Include the "why" for non-obvious answers
- Organising your knowledge base
- Measuring effectiveness
- Setting up in Kerabie
Most knowledge bases are written by someone who already knows the product — which is exactly why they fail to help customers who don't. Here's how to build articles that actually deflect support tickets.
The knowledge base reality check
Before writing a single article, pull your last 100 support conversations and categorise the questions. You'll find that 20–30 questions account for 70–80% of your volume. Those are your first articles.
Writing articles that work
Use the exact words your customers use
Search your support conversations for the exact phrases people type. If customers say "I can't log in" rather than "authentication failure", your article title should be "I can't log in to my account."
Answer the question in the first paragraph
Don't make people scroll to find the answer. Lead with it. Context and detail come after.
Bad: "Kerabie's password reset feature is part of our authentication system which..."
Good: "To reset your password: click Forgot password on the login page, enter your email, and check your inbox for a reset link."
Use numbered steps for processes
Every action that requires a sequence should use numbered steps. Readers scan, not read. Numbers make scanning easy.
Screenshots for anything visual
If the answer involves clicking something in a UI, show a screenshot. Text descriptions of interfaces are universally confusing.
Include the "why" for non-obvious answers
If the answer seems strange ("You can't export before 24 hours"), explain why. It prevents follow-up questions and frustration.
Organising your knowledge base
Group articles into 4–6 categories maximum. More than that and customers can't navigate.
Common structures that work:
- Getting started → Account settings → Billing → Integrations → Troubleshooting
- By feature → By role → Troubleshooting
Measuring effectiveness
Track:
- Deflection rate — % of people who read an article and don't start a chat
- Article views — which articles get the most traffic
- Helpful/Not helpful votes — which articles need rewriting
- Search terms with no results — gaps in your coverage
Setting up in Kerabie
Kerabie's AI assistant is trained on your knowledge base automatically. Every article you publish becomes part of the AI's knowledge — so the better your articles, the better your AI responses.