A person may know exactly which document they want, or only the question they need to answer. Those are different discovery tasks. Search and navigation should accommodate both without asking visitors to learn the organization’s internal filing system.
Understand what is being searched.
A search interface inherits the quality of its source material. Duplicate documents, missing titles, outdated records, and ambiguous labels can create poor results even when the search technology is capable. Reviewing a representative sample helps locate the problem before choosing a tool.
We frame search as a relationship between a collection and an audience. Useful planning considers which sources belong in the index, how quickly changes must appear, and which information each audience is allowed to see.
Design more than one route.
A keyword query is one route into a collection. Topic lists, filters, related items, and carefully named categories can provide others. The right combination depends on the collection’s size and the way people describe their needs.
Potential work includes result-page design, metadata improvements, search integration, filter behavior, synonym planning, and empty-result guidance. A result should help a person judge relevance before opening the item: a descriptive title, useful excerpt, content type, and freshness indicator can matter more than extra decoration.
Review the difficult queries.
A practical evaluation set contains known-item searches, broad exploratory questions, ambiguous terms, and queries that should produce no results. Recording the expected behavior turns a vague discussion about relevance into a reviewable task.
- Identify a small set of high-value visitor questions.
- Check permission boundaries and unavailable content.
- Make empty and error states useful.
- Record changes that improve or weaken relevance.
Choose an approach with visible limits.
Some collections need a well-organized directory rather than a new search service. Others benefit from indexing or more advanced retrieval. The first recommendation should reflect the actual collection and maintenance capacity. No search interface should imply that it covers sources it cannot access.
Our search guide describes how a small evaluation collection can support better decisions without turning the project into a research program.
Read the search evaluation guide