itsual Search Design and Indexing
Overview
Search is a product feature as much as a technical one. Relevance, latency, and freshness all matter.
When to Use
- Adding or redesigning search
- Choosing between database LIKE/full-text, dedicated search engines, or vector search
- Improving relevance or performance of existing search
Key Decisions
- What needs to be searchable (fields, entities, permissions)?
- Indexing strategy and freshness requirements
- Ranking / relevance signals
- Filtering, faceting, and sorting needs
- Handling of permissions and multi-tenant isolation in results
- Typo tolerance, stemming, language support
Principles
- Start with the user task, not the technology
- Measure search quality (relevance, zero-result rate, latency)
- Keep indexing pipelines observable and recoverable
Verification
- Common user queries return relevant results
- Permissions are enforced on search results
- Indexing lag and failures are visible