With 56,000 games on Gameziko, search is easy. Discovery is the hard part. Nobody types what they want when they don't know it exists. Here's what we learned building categories people actually browse.
The paradox of choice is real
Our first category page was an alphabetical list. It tested terribly. Users scrolled, got overwhelmed, and left. The fix wasn't fewer games, it was better doors into the catalog.
We now treat categories as answers to moods, not genres. "Puzzle" is a genre. "Games you can finish on a coffee break" is a mood. Mood-based groupings consistently outperform genre lists in click-through, sometimes 3 to 1.
What the data showed
A few patterns from our analytics:
- Session length beats game count. Categories with 50 well-chosen games outperform categories with 500 unsorted ones. Curation is a feature.
- The second click matters most. If a user clicks a second game, session length triples. So every game page ends with "if you liked this" recommendations, not a generic footer.
- New users browse, returning users search. First-time visitors use categories 4x more than search. Search only wins after someone knows what they like.
How we built it (the practical bits)
Our category pages are statically generated from a tagging pipeline. Each game gets:
- 2-4 mechanic tags (match-3, physics, tower-defense)
- 1 session-length tag (quick, medium, long)
- 1 mood tag (relaxing, competitive, silly)
The pipeline runs nightly over the full catalog. Tags are 90% heuristic (title keywords, play-time data, engagement signals) and 10% hand-tuned for the top 500 games, because the head of the catalog deserves human judgment.
Rendering is the boring part done right: static HTML, no client-side filtering for the initial view. Filter chips enhance progressively. A category page with 200 game cards loads in under a second because there's no framework booting up to render a list.
The one thing I'd do differently
We built "related games" from shared tags first. It was fine. Then we switched to shared play patterns (players who finished game A also finished game B) and engagement jumped. Tags describe games; behavior describes players. Use both, but weight behavior higher.
If you're building any kind of catalog, start with moods, curate the head by hand, and measure the second click. Everything else is optimization.
I run Gameziko, a free browser games portal. No downloads, no sign-ups.
