A customer searches a fashion shop for "sneakers." The shop carries exactly the pair they want, but the product text refers to it solely as "trainers." The result: an empty results list, even though the item is in the warehouse, correctly priced and available for immediate delivery.1 From the customer's perspective, this looks like a gap in the product range. In reality it almost never is – it's a translation problem between the language customers use and the language of the product catalog.
That distinction is rarely made in most shops. Internally, a zero-result search is often read as evidence of a missing product, when in the overwhelming majority of cases it's a sign that the search logic simply failed to interpret the query correctly. Anyone who misses this distinction optimizes the wrong lever – and keeps losing revenue to a problem that is entirely solvable.
The anatomy of a zero-result search
Just how widespread this misunderstanding is becomes clear in an analysis by the Baymard Institute: 56 percent of the online shops examined fail to adequately support their users' search needs.2 The problem almost never lies with the search engine itself, but with its configuration: depending on the settings, a multi-word query like "blue running shoes" returns either thousands of barely filtered results or none at all, German compound words break apart uncontrollably, and product codes get lost among strings of digits.2 These causes fall into four recurring patterns that together account for the bulk of all zero-result searches.
Cause 1: Typos and spelling variants
The most obvious – but by no means the smallest – cause is simply mistyped queries. According to an analysis by Wizzy, typos and spelling variants alone account for 5 to 15 percent of all failed searches.1 A real-world example shows how much difference good versus poor error tolerance makes: at the brand Burt's Bees Baby, a misspelled query for "pyjamas" still reliably leads to the matching product "pajamas," because the search recognizes the similarity between the two spellings.3 Many shops fail to do exactly that – a single transposed or missing keystroke is enough to produce an empty results page.
Cause 2: Missing synonyms
A second problem weighs even heavier: missing synonyms. In broad product catalogs, missing synonym mappings account for up to 30 percent of all unsuccessful queries, according to the same analysis.1 The "sneakers" example above is not an exception but the rule: customers often use completely different, colloquial terms for one and the same product, while the product text usually contains only a single official designation. A real B2B example shows how effectively this problem can be addressed: the tool and fastening technology supplier Berner deliberately created synonym entries for colloquial search terms and was able to significantly reduce the number of zero-result pages in its German- and French-language shops.4
Cause 3: German compound words – a homegrown problem
One problem that is particularly often overlooked in German-speaking commerce concerns the language itself: German queries frequently consist of compound words that don't exist in this form in English. When a customer searches for "Damenhose" (women's trousers), the search engine has to be able to technically split this word into "Damen" and "Hose" in order to match it correctly against the product index – a process based on a dictionary of the most common German terms that varies considerably from one search solution to the next.5 Similar problems arise with hyphenated spellings: if someone searches for "Wander-Schuhe" with a hyphen, the search additionally has to recognize that this carries the same meaning as the single-word variant "Wanderschuhe."6 Shops whose search index doesn't reliably handle this decomposition produce a zero-result page for every one of these phrasings – regardless of how well the actual product would have matched the query.
Cause 4: The invisible configuration trap
Perhaps the least well-known but most consequential technical cause is buried deep in the search logic itself: the parameter that determines how many words of a multi-word query a product must match in order to appear in the results list at all. If this setting isn't deliberately configured, a search often works in practice either purely additively – so that almost every product containing just one of the search terms appears – or, conversely, too strictly, so that a product has to contain all search terms exactly and drops out of the results at the slightest deviation.2 Both are frustrating for customers: in the first case, the product they want is buried under a flood of irrelevant results; in the second, it disappears entirely even though it matched the intent of the query precisely. This misconfiguration often goes unnoticed for years, because it doesn't feel like an obvious error – it feels like an inconspicuous but constant drag on revenue.
Making matters worse, these default settings are rarely intuitive and differ from one search system to the next. Anyone who puts a system into operation without deliberate adjustment often runs their search effectively on pure OR logic without ever having explicitly decided to – a configuration detail that hardly ever appears in the requirements document for a shop relaunch, yet decides every day whether revenue is found or lost.
The fallacy: zero results are rarely an assortment problem
All these causes lead to the same expensive consequence. The Baymard Institute puts the average zero-result rate in online shops at 8 to 15 percent – and each of these zero-result searches reduces purchase probability by as much as 88 percent.7 What makes it especially painful: these are precisely the customers with above-average purchase intent, actively typing because they want to buy. The fallacy in many companies is to read a high zero-result rate as a sign of too small a product range and to invest in new products instead of better search logic. In the vast majority of cases, a look at your own search logs tells a different story: the product being searched for was in the catalog the whole time – just under a different word, in a different spelling, or hidden behind overly strict query logic.
This fallacy also has an organizational cause: zero-result statistics usually land in the IT department, while decisions about expanding the product range are made in purchasing. Direct exchange between the two departments is often missing, so a technical search problem ends up reaching purchasing as a supposed assortment gap – even though no one there can do anything about it. Sharing zero-result data regularly between search owners, category management, and product data maintenance substantially narrows exactly this blind spot.
How Eywora structurally prevents zero-result searches
Eywora's semantic AI search addresses these four causes deliberately and together, rather than treating them as separate individual problems. Typos and spelling variants are detected and corrected automatically; synonyms aren't maintained manually in a rigid list but understood semantically, so that a query for "sneakers" reliably also finds products listed in the catalog as "trainers." German compound words are correctly decomposed and reassembled, so that both "Damenhose" and "Damen Hose" lead to the same matching result. And instead of rigidly configured AND/OR logic, a reranking model sorts results by their actual relevance to the query rather than filtering products solely by how many individual words matched exactly.
To keep the problem from quietly returning, the analytics component continuously evaluates which queries still lead to few or no results and prioritizes them by their estimated revenue potential – so teams can start where a fix will have the greatest effect.
If you want to know how many queries in your own shop currently fail for exactly these four reasons, you can test Eywora on your own catalog free for 14 days without a credit card, or book a 30-minute live demo using a real customer shop.
A practical roadmap: how to identify your biggest sources of zero results
The first step requires no new technology, just an honest look at your own search logs. Export the most frequent queries that return zero or conspicuously few results and assign each one to one of the four causes described: typo, missing synonym, unresolved compound word, or overly strict query logic. In practice, the problem usually concentrates in a handful of patterns that repeat again and again – once you've identified them, targeted corrections can eliminate a disproportionately large share of all zero-result searches, long before any larger technical overhaul becomes necessary.
Conclusion: the customer didn't search for the wrong product
The most important insight from all these causes is a simple shift in perspective: when a customer can't find a product that's in the catalog, it's rarely the customer who did something wrong. They simply spoke a different language than the product catalog. Shops that understand this distinction and adapt their search logic accordingly turn one of the most expensive – and simultaneously most invisible – weak points in the entire purchase process into one of the fastest to fix.
Sources
- XICTRON – Site search analytics: putting in-shop search behavior to use, citing Wizzy – https://www.xictron.com/de/blog/site-search-analytics-suchverhalten-online-shop ↩
- XICTRON – Search relevance in the shop: matches instead of result floods, citing the Baymard Institute, Shopware AG, and the OpenSearch documentation – https://www.xictron.com/de/blog/suchrelevanz-shop-opensearch-tuning-2026 ↩
- Nosto – Search on e-commerce websites: best practices and examples – https://www.nosto.com/de/blog/suche-auf-e-commerce-websites-bewaehrte-verfahren-und-beispiele/ ↩
- ap-verlag.de – B2B e-commerce – the search function as a "flagship" – https://ap-verlag.de/b2b-e-commerce-die-suchfunktion-als-aushaengeschild/72217/ ↩
- Tudock – E-commerce onsite search basics #1: how it works – https://www.tudock.de/archiv/grundwissen-ecommerce-onsite-search-1-funktionsweise/ ↩
- FACT-FINDER – Intelligent search: how your online shop understands every request – https://www.fact-finder.de/blog/de/intelligente-suche/ ↩
- Doofinder – Site search statistics 2026: figures on search in online shops, citing the Baymard Institute – https://www.doofinder.com/de/blog/e-commerce-statistik ↩