Relevant, or recent? How to make hybrid search work for news.

Search “interest rates” on a news site and you want this week’s story, not the best match from 2019. At Het Financieele Dagblad, we built hybrid search across an archive of more than a million articles. Hybrid search combines traditional keyword matching with semantic embeddings, promising better retrieval accuracy. At first, it delivered. However, standard hybrid search falls short in news, where search results must balance semantic relevance with article freshness. We will walk through how hybrid search works, why standard recency adjustments failed us, and the unexpected traps we encountered while trying to fix them. Ultimately, solving this required us to rethink not just our ranking algorithms, but how we measured search quality altogether. Takeaways: You’ll leave with a clear intuition for how hybrid search works, how to evaluate search quality effectively, and practical techniques to tune such ranking systems when the default frameworks fall short.