Building Your Product Taxonomy — Example & Best Practices
Your product taxonomy is a structured hierarchy that BuildBetter uses to automatically categorize signals from calls, feedback, and other sources. A well-built taxonomy ensures every piece of customer feedback gets routed to the right product area — so nothing falls through the cracks.How it works
- Paste your product documentation (feature lists, help docs, marketing pages, release notes) into the taxonomy generator
- AI extracts a 4-level hierarchy from your documentation
- Signals get auto-labeled against your taxonomy as they come in
The 4-level hierarchy
Domains organize your product at the highest level. Products are the modules within each domain. Features are the things users actually interact with. Tags are the specific labels AI applies to signals.
Only Product, Feature, and Tag levels are auto-labeled. Domains provide organizational context but are not directly applied to signals.
Example: Project management platform
Below is a complete taxonomy for a fictional project management tool called “ProjectFlow.” This is the kind of document you’d paste into the taxonomy generator.Input document
Generated taxonomy
Here’s what the AI produces from the above input — shown as the tree view you’ll see in BuildBetter’s settings:What makes this taxonomy effective
1. Domains reflect how customers think, not how your org is structured
“Work Management” and “Collaboration” map to customer jobs-to-be-done, not internal team names like “Core Platform Team” or “Growth Squad.”2. Products are distinct modules with clear boundaries
Each product (Task Boards, Documents, Messaging) is something a customer would recognize as a separate thing. Avoid overlapping products — if two products share features, pick the primary home.3. Features are specific and actionable
“Comments & Threads” is better than “Social Features.” “Sprint Planning” is better than “Agile.” Name features the way your customers and your docs name them.4. Tags have clear instructions
Every tag includes aninstructions field that tells the AI exactly when to apply it. This is the most important part — vague instructions lead to inaccurate labeling.
Good instruction: “Apply when the signal mentions creating tasks from Slack or receiving Slack notifications”
Bad instruction: “Apply when related to Slack”
5. Tags capture specific scenarios, not duplicates of the feature name
Tags should represent variations, use cases, or specific aspects — not just restate the feature. If a feature only has one scenario, one tag is fine.Tips for writing input documentation
The quality of your taxonomy depends on the quality of the text you provide. Here’s what to include:
Best sources to paste in:
- Help center articles or knowledge base
- Feature comparison pages
- Release notes (last 6-12 months)
- Product marketing pages
- API documentation overviews
- Security/compliance pages
- Internal jargon customers wouldn’t use
- Org charts or team structures (these aren’t product areas)
- Extremely long documents with redundant content (the generator has a 100K character limit)
After generation: refine and iterate
The generated taxonomy is a starting point. After generation:- Review the tree in Settings > Features > AI Labeling — rename or remove anything that doesn’t match your product vocabulary
- Add missing tags for common support scenarios the docs didn’t cover
- Edit tag instructions to be more specific to your customer base
- Re-generate anytime your product significantly changes — paste updated docs and the taxonomy will be replaced with a fresh structure