The zero results page: search demand your store collected and then dropped
Read the log like a worklist, not a report
Pull the zero results queries weekly and sort by frequency. The top of the list is almost always fixable in an afternoon: plural and singular forms, spacing and hyphen variants, brand names with the words in the wrong order, and the plain vocabulary gaps where customers say couch and the catalogue says sofa. Each one is a shopper who arrived with intent and got a closed door, which makes this the highest return hour in site search work.
The tail of the log teaches different lessons. Queries for products you do not sell, repeated often, are either a range gap worth costing or a sign your ads and social posts are promising something the store never stocked. Queries that are really questions, sizing help, delivery times, belong routed to content, and a zero results page that suggests the right guide page keeps that session alive.
Synonyms are merchandising, not IT
Somebody who knows the products has to own the synonym list. Search tools ship with generic dictionaries that know nothing about your trade's slang, your brands' alternate names, or the words your customers actually use, which support tickets and reviews will tell you for free. Add terms in pairs, test them against real queries, and prune the ones that pollute results by matching everything. A synonym list that makes every search return forty products is a different failure wearing the same costume.
Misspellings deserve the same respect. A shopper typing on a phone with one thumb is not a careless person, they are your median user. Fuzzy matching tuned too tight returns nothing for a one letter error. Tuned too loose, it returns confident nonsense. Test with your own top queries misspelled the ways the log shows them misspelled, not the ways a test script invents.
The page itself should sell the recovery
When a search truly has no match, the page has three jobs: say so in one line without blaming the shopper, offer the nearest useful alternatives, and capture the demand. Nearest alternatives means the closest category, the best sellers in the adjacent range, or the products matching the loosest sensible reading of the query, labelled honestly as alternatives rather than dressed up as results.
Never auto redirect a zero results query to a category page without telling the shopper. The redirect hides the failure from your analytics, confuses the shopper who typed something specific, and teaches the team the problem is smaller than it is. A labelled fallback keeps the data honest and the shopper oriented, and honesty in the data is what gets next quarter's fix funded.
Track zero results rate as a share of searches, the exit rate from those pages, and conversion on sessions that hit a fallback and stayed. The first number should fall as synonyms land. The second tells you whether the recovery page works. Judge a synonym change by the queries it targeted, not by overall search conversion, which moves with traffic mix and will happily take credit for your work or blame you for the weather.
Measure the fix where it happened
We shipped this review loop on stores where search drives a third of revenue and the first pass over the log, an hour with the top fifty dead queries, recovered more sessions than a month of debating the search vendor. The demand was already there. It was spelling itself differently.
Frequently asked questions
How often should we review zero results queries?
Weekly for an active store, with the top queries by volume actioned each time. The list changes with seasons, campaigns, and new ranges, so a one off cleanup decays within a quarter.
How often should we review zero results queries?
Yes, as clearly labelled alternatives, ideally from the closest category to the query. Random best sellers from across the store read as a shrug. Relevant neighbours read as help.
How often should we review zero results queries?
Sometimes, but vocabulary gaps and catalogue naming cause more zero results than engine quality. Fix the synonyms and naming first. They are cheaper than a migration and they transfer if you do migrate.
How often should we review zero results queries?
There is no universal benchmark worth trusting. Track your own trend. A rate falling after each review cycle, with exits from those pages falling too, is the signal that matters for your store.