Site search implementation.
Search that finds things — including when people spell them wrong.
Site search is frequently the most-used and least-invested-in feature on a store or content site. People who search convert at much higher rates than people who browse, and a search that returns nothing for a near-miss spelling sends them straight to a competitor. Database LIKE queries do not solve this.
What’s included
- Full-text search with typo tolerance
- Faceted filtering that stays fast on large catalogues
- Synonyms and custom relevance rules
- Autocomplete and search-as-you-type
- Result ranking you can tune commercially
- Analytics on what people search and what returns nothing
- Multi-language support where needed
Zero-result searches are a to-do list
Every search returning nothing is either a product you should stock, a synonym you should add, or a spelling you should tolerate. Logging and reviewing them is one of the highest-return, lowest-effort activities available on a store.
Relevance is commercial as well as technical
The most textually relevant result is not always the one you want ranked first. In-stock items, higher-margin products and seasonal lines may deserve a boost. That should be a tunable business rule, not something buried in code.
Common questions
Which search technology do you use?
Meilisearch or Typesense for most sites — fast and simple to run. Elasticsearch for large or complex needs. Algolia when you would rather not host it. I will recommend based on catalogue size and budget.
Can it handle filters on a large catalogue?
Yes — faceted filtering is where a proper search engine dramatically outperforms database queries.
Can we control what ranks first?
Yes, with boosting rules and pinned results for specific queries.
Have a project in mind?
Tell me what you're building and roughly when you need it. I reply to every genuine enquiry within one working day.