Synonyms are the cheapest, highest-impact relevance fix in ecommerce search. A two-line rule that maps “couch” to “sofa” recovers a measurable chunk of zero-result and low-conversion queries overnight, with no model retraining.
They come in three flavors: bidirectional (couch ↔ sofa, both queries return both kinds of results), one-way (snickers → sneakers, but not the reverse), and hierarchical (footwear → sneakers, sandals, boots — used for category expansion). Most search engines support all three.
Auto-detection is the next step beyond hand-curated lists. By mining co-clicked products from search logs, or using embedding clustering on product titles, you can surface synonym candidates that a human reviewer can approve. Stores that automate this catch hundreds of synonyms a human would never think to add.