Zero Results Pages: Where Search Relevance Goes to Die

A zero results page means a shopper told you exactly what they wanted and you answered with nothing. Most are caused by vocabulary mismatch, over strict filters, and catalogue data gaps, not by genuinely missing products. Fixing zero results queries is usually the highest return search work a store can do.

The most honest report in your analytics

Every zero results query is a shopper raising their hand. They named a product, a use case, or a problem, and the catalogue either lacks it or the search failed to connect their words to your words. Unlike bounce rate or vague engagement metrics, zero results data is unambiguous: here is demand, expressed in the customer's language, that you did not meet.

For more on this, see this related guide and our earlier explainer on the same theme.

Start by splitting the report in two. Queries for things you do not sell are assortment intelligence; log them for buying decisions and move on. Queries for things you do sell are relevance failures, and they are embarrassing in direct proportion to how obvious the match is. A store that sells trail running shoes and returns nothing for off road trainers has a vocabulary problem, not a catalogue problem.

Vocabulary mismatch and the synonym gap

Customers search in their language, catalogues are written in the merchant's. Sofa versus couch, trainers versus sneakers, and every industry's jargon versus plain speech create gaps that keyword matching cannot cross. Synonym dictionaries close the known gaps, but the durable fix is semantic: search that understands intent and attributes rather than literal tokens.

Mine the zero results report weekly and feed it back. Add the synonyms, fix the product titles missing the everyday term, and correct the attribute data that made a real product invisible. Each fix is small; the compound effect is a search that speaks the customer's language a little better every week, and the report shrinks accordingly.

Filters and strictness: the self inflicted zero

A shocking share of zero results come from the shopper's own filters carried too far: a size, colour, and price combination that empties the set, remembered silently from an earlier page. The search is technically correct and commercially useless. When filters produce zero, the page should say which filter emptied the results and offer one click removal, not present a bare apology.

Readers also found this companion piece from our archive useful.

Token strictness does the same damage invisibly. Requiring every word in a long query to match guarantees failure on natural language searches like warm waterproof jacket for winter hiking. Relevance ranking should interpret long queries as weighted wishes, not as a contract where one unmatched word voids the whole search.

Designing the recovery, not the apology

Some zero results are real: the product is out of stock or genuinely not carried. That page still has a job. Show the closest matches with an honest label, offer the category path the query implies, and capture the demand with a notify me or a search alert. A shopper who leaves a dead end with an alert set is a future sale; a shopper who hits a blank page is a bounce statistic.

Track recovery rate as the metric: the share of zero results sessions that go on to click something or buy. It converts a depressing failure report into an optimisation target, and it keeps the team honest about the difference between fixing search and decorating the apology page.

Frequently asked questions

What causes most zero results pages?

Vocabulary mismatch between customer language and catalogue language, filters silently carried into the search, and overly strict matching that requires every query word to match. Genuinely missing products are usually the smaller share.

How often should zero results be reviewed?

Weekly for active stores. Split the report into products you do not sell, which informs buying, and products you do sell, which are relevance bugs to fix with synonyms, titles, and attribute corrections.

What should a zero results page show?

An honest explanation, which filter to remove if filters caused it, closest matches clearly labelled, the implied category path, and a way to capture demand such as a stock alert. Never a bare apology with no next step.

What metric tracks improvement?

Recovery rate: the share of zero results sessions that continue to a click, an alert signup, or a purchase. Pair it with the falling count of zero results on products you actually sell.