Rebuilding Product Hunt's Taxonomy to Drive Organic Growth
Product Hunt's taxonomy was a flat, Tumblr-style tag system. "iPhone App" alongside "Artificial Intelligence" alongside "Cats." It didn't map to how people actually searched for products, and competitors were ranking for "best [category] tools" while we mostly ranked for specific product names.
I replaced it with a two-tier category system, manually categorized 5,000+ products to avoid a cold start, and designed thousands of programmatic SEO landing pages to capture discovery traffic.
Hundreds of new keywords ranked from category pages capturing broad discovery searches
Tens of thousands of monthly organic visits from programmatic landing pages
6,000+ makers self-categorized within weeks of launch
Problems
Inconsistent Taxonomy
The flat tag system mixed categories, platforms, technologies, and memes. No hierarchy or logical grouping made browsing nearly useless.
Missing Discovery Traffic
Organic searches researching products that solved specific problems weren't finding Product Hunt. Competitors ranked for "best [category] apps" while we ranked mostly for specific product names. Good, not great.
Cold Start Risks
Launching a new category system without pre-populated data would result in empty pages.
Approach
I started by auditing the entire existing topic system, classifying each tag by what it actually represented: platform ("iPhone App"), technology ("Artificial Intelligence"), feature ("Dark Mode"), category ("Project Management"), or just noise ("Cats"). That audit, combined with internal search data, revealed that users weren't searching for topics the way we'd structured them. They were looking for product categories like "note taking tools" or "CRM software," which we didn't have. I cross-referenced with product comparison sites to design a new, two-tier category system with parent and child subcategories.
Because these categories didn't exist yet, we had a cold start problem: launching with empty pages would be worse than the old system. I manually reviewed and categorized over 5,000 popular products before launch and developed internal guidelines for category definitions to keep things consistent going forward.
Keyword research revealed the SEO opportunity. There was substantial search volume for "best [category] tools," "[category] apps," and feature-specific queries — traffic competitors were capturing that we weren't. This shaped the decision to pair the taxonomy work with programmatic landing pages.
Solution
The system launched with a two-tier hierarchy where parent categories contained relevant child categories. Each page featured top products, recent launches, filtering options, and descriptive content. No empty states thanks to the manual categorization work.
On top of this, thousands of programmatically generated landing pages captured long-tail traffic. Category pages covered broad topics like "AI noteakers" and "Headless CMS Software," while topics remained but were deprecated in importance. Every category page was genuinely useful, not SEO bait.
We added category tagging to new product launches (limited to 2), keeping content fresh as more products were added.
Impacts
Significant Organic Growth
Programmatic pages captured new organic traffic, ranking for hundreds of category search terms and driving tens of thousands of monthly visits.
6,000+ Maker Adoptions
Over 6,000 makers categorized their products in the first few weeks after launch, validating that the system was intuitive and valuable.
Reflections
Manually categorizing 5,000 products was tedious but absolutely necessary. The upfront investment paid off immediately. We had no empty states at launch, and the quality seed data gave the programmatic SEO strategy a strong foundation to build on. Plus, any products that were not already categorized had an incentive to do so to show up with their competitors.
Another surprise about the old topics system was discovering that we had an unfortunate feedback loop. The Topics page was sorted by topic popularity, but it turned out that this was just being reinforced by our selection UI when adding new products, which was also sorted that way. So we ended up with extremely general, not useful topics as the most used.