Customers rarely search with the same vocabulary used in a business's product database. A maintenance customer may remember a street, a month or part of a job description. A trade buyer may know a size but not a product code. Search on a small screen should help people narrow uncertainty, with enough context in each result to recognise the right item.
Learn the words customers actually use
Collect common terms from service calls and observe a few people trying to find real items. A hardware catalogue may need both trade terminology and ordinary descriptions for the same product. Decide which attributes should be searchable and which should become filters. Avoid silently changing a query in a way that surprises the customer. Show useful matching details, such as branch, size or appointment date, so similar results can be distinguished without repeatedly opening and closing them. Keep the entered query visible while the person explores.
Make filters explain the result set
A customer filtering appointments by location should be able to see that the filter is active and remove it easily. Hidden filters can make a working search look broken. Use clear counts and descriptive selections where helpful, and preserve the state when returning from a result. Apple's searching guidance provides a platform design reference; the business still needs to decide the vocabulary and useful result content. On a phone, prioritise a small number of meaningful filters rather than placing every database field in a long panel.
Recover when nothing matches
Distinguish no matching items from unavailable search due to a connection problem. Offer a useful next action: remove a restrictive filter, try a related term or browse a relevant category. Avoid blaming the user with an unexplained No results label. Test misspellings, partial identifiers, long queries and several similar records. Also test the keyboard covering the result area and the screen reader announcing updated results. Measure whether customers find the correct item, not merely whether the search returns something. An irrelevant first result can be more costly than an honest empty state.
- Which customer words differ from the labels stored in the business system?
- Can someone tell why a result matched without opening every item?
- Are active filters visible when the result list becomes unexpectedly short?
- Does the empty state distinguish no matches from a temporary search failure?