What are Related Searches?
Related searches are query suggestions generated from user behavior and content signals that are semantically or behaviorally close to the original query. They reduce dead ends and speed up discovery by offering next-best intents.
How It Works (quick)
- Signals: Co-typed/co-clicked queries, session paths, co-purchased items, embedding similarity, knowledge-graph neighbors, and taxonomy proximity.
- Generation: Mine logs, cluster by intent, rank by likelihood + diversity, localize by market.
- Placement: Above results, after first row, or at the bottom; show 4–8 concise options.
- Personalization (optional): Reorder with user/segment affinities (privacy-aware).
- Safety: Filter to indexed, in-stock, policy-compliant destinations.
Why It Matters in E-commerce
- Cuts zero-results and pogo-sticking.
- Captures long-tail demand and related missions (e.g., care guides, size charts).
- Improves internal linking to high-value collections and content.
Best Practices
- Mix narrower (add facets/brands) and broader (parent category) options; avoid near duplicates.
- Use clean, human phrasing (no raw operators/IDs).
- Respect locale (sizes, currency, synonyms).
- Track usage, CTR, reformulations, assisted revenue; prune low performers quarterly.
- Don’t loop: suggestions shouldn’t just restate the same query.
Challenges
- Spammy or biased suggestions from noisy logs; seasonal drift; multilingual ambiguity.
Examples
- “trail shoes” → trail running shoes waterproof, trail shoes men size 45, gaiters, trail socks.
- “iphone charger” → USB-C PD 30W, MagSafe charger, charging cable 2 m.
Summary
Related searches suggest the next best intents. Generate from logs and semantics, enforce safety, localize phrasing, and measure assisted success.
