Elasticsearch Search Quality & Module Consolidation
Challenge
A flagship product was appearing at position 3 in search results when customers searched for its exact product name, with two less relevant products ranked above it. The e-commerce team had escalated the issue as it was directly affecting conversion on one of their top-selling SKUs.
What Was Built
Root cause: individual products had been assigned custom Elasticsearch search weights (product_search_weight attribute) during a past configuration exercise, causing competing products to score higher than exact-match products for the same query. Rather than patching the three affected custom Elasticsearch integration modules in isolation, the overlapping logic was consolidated into a clean single module, eliminating duplicated overrides. Fixed sort order for both search results and autocomplete suggestions. Multiple search query manipulations in the Elasticsearch client to improve relevance scoring overall.
Outcome
Exact-match product search returns the correct product at position 1 across tested queries. Autocomplete suggestions follow the same improved ranking. The module consolidation reduces future maintenance surface for all Elasticsearch customisations.
"Tests look good - please bring it live."
Have a similar challenge?
Get in touch - no sales pitch, just a straightforward conversation.