Revamping In-App Search: A UX Case Study Using Vector Search at Bookory

AI · 6 min read

Bookory, a reading discovery app, struggled with users failing to find niche or author-contextual results using traditional keyword search. The team introduced a hybrid approach: semantic vector ranking for intent matching plus deterministic filters for format, language, and publication date. Designers focused on surfacing explainability signals so users understood why a recommendation matched their query.

The UI showed a highlighted rationale card above results—for example, 'matches your interest in modern sci-fi themes'—and allowed users to fine-tune results with inline chips. The underlying design trade-offs included deciding when to favor semantic matches over exact-title matches and how to let users revert to exact-match mode quickly. The team also created a feedback affordance so users could mark a result as irrelevant and teach the ranking model.

After the revamp, relevance metrics improved: click-through rate on first-page results rose 24% and time-to-first-click dropped 18%. Engagement for long-tail titles increased, and labeled feedback accelerated re-ranking on common misfires. The article closes with guidance for designers integrating vector search: provide transparency, offer deterministic fallbacks, and close the loop with user feedback to keep relevance aligned with expectations.