Search performance with 50k+ SKUs and 30+ attributes — how fast is fast?

Target: <200ms search response time at p95, <100ms layered-nav filter response. Achievable with:

• Elasticsearch / OpenSearch: required since Magento 2.4. Run a dedicated cluster (not co-located with MariaDB) at 4–8GB RAM minimum, NVMe storage. Single-node is fine to ~250k SKUs; shard to 3 nodes above that.
• Searchable attribute curation: not every attribute should be searchable. Review the catalog_eav_attribute.is_searchable flag — turn off "Cost", "Width Map" etc. that customers don’t search. Fewer searchable attributes = faster index + faster query.
• Layered nav indexing: catalog_product_attribute + catalog_search indexers on cron, not on save. Re-index 50k SKUs in 8–12 minutes nightly.
• Synonyms + boost rules: configure synonyms (e.g. "WiFi" / "Wi-Fi" / "wireless networking") in admin → Stores → Search Synonyms. Boost rules promote in-stock + high-margin SKUs.
• Hyvä autocomplete: replace Magento’s default suggest with Hyvä-native autocomplete — ~200ms keystroke latency vs Luma’s 600ms+.
• Optional: Algolia / Klevu / Searchspring — SaaS search overlay. $200–$2,500/mo. Worth it at >100k SKUs or when on-site search is a primary conversion driver. They handle relevance tuning, NLU, personalization out of the box.

For a typical 50k-SKU electronics store, vanilla Magento + Elasticsearch + Hyvä is enough. Add Algolia/Klevu only if the data shows search is a top-3 conversion driver.