The catalogue has the product but search denies it
A customer searches using the name printed on a replacement part. The catalogue contains the compatible item under an internal description, and the result page reports nothing found. The search engine has answered its query correctly while the product has failed the customer's task.
An empty result is evidence to investigate, not proof that the catalogue lacks an answer. The failure may sit in terminology, indexing, filters, availability or the assumptions made about what the customer wants. Changing the empty-state illustration will not distinguish those causes.
Begin with the request and the intended outcome. A replacement part search, a category browse and an exact model lookup can contain similar words while requiring different recovery behaviour.
Separate absence from exclusion
For a hypothetical equipment store, search the same request with and without the selected filters. If the desired item appears only after a delivery filter is removed, the catalogue is not empty. The current set of conditions excludes the product.
The interface should say that and offer a deliberate change. Show which condition would be relaxed and what the alternative means. Quietly removing a compatibility or delivery constraint risks returning a product that looks relevant but cannot serve the customer's purpose.
Elastic's search documentation distinguishes full-text search from structured filtering. Its query guidance covers those mechanisms. The product decision is how those mechanisms interact with the customer's non-negotiable requirements, not merely which query returns more rows.
Treat vocabulary as catalogue work
The names customers use may differ from supplier labels and internal identifiers. Collect representative failed searches and review them with someone who understands the merchandise. Determine whether the request names a real product, a common alternative term or an incompatible item.
Add language support deliberately. A synonym should express a relationship the business is willing to stand behind. Similar looking terms can identify different sizes, standards or intended uses. Turning them into interchangeable words can trade an empty result for a wrong purchase.
Product data often needs the repair. Put recognised identifiers, compatibility descriptions and useful attributes into the catalogue rather than hiding all knowledge inside search rules. That makes the information available to product pages, support and future channels as well.
Offer a recovery that preserves intent
When the request remains ambiguous, ask a useful clarifying question. For the equipment example, the missing distinction might be the device family or connection type. A generic try another keyword message transfers the whole diagnosis back to the customer.
If the product truly is not offered, say so. Alternatives should explain the relevant difference instead of disguising unrelated stock as a match. A supported enquiry route can be more helpful than confidently filling the page with products that cannot satisfy the request.
Keep the original search and active conditions visible during recovery. The customer should understand how the new result differs from the first attempt and be able to return without rebuilding the entire query.
Review failed tasks this week
Take recent unsuccessful searches from appropriately collected records and group them by likely cause. Avoid storing unnecessary personal text. Choose examples that represent terminology gaps, overrestrictive filters, unavailable products and genuine catalogue absence.
For each example, write the expected response with a merchandiser or support colleague. Implement the smallest change that makes that response possible, then check that previously correct searches still work. Track whether the customer reaches a suitable product or an honest next step. A fuller result page is not a success if it only makes the wrong answer easier to buy.
