We analyzed search logs from more than 50 B2B shops across manufacturing, wholesale and spare parts. The reality is uncomfortable: on average 38% of all search queries return zero or unusable results. A significant share of those come from customers who actually wanted to buy.
Three patterns we see again and again
1. Standard search engines don’t understand compound words
“High-pressure hose DN10 1SN” gets split by a typical shop search into three words, none of which appear in the product title. The search returns zero, even though the matching item is in stock. With semantic decomposition, the same result would appear in 47 ms.
2. Material codes get read as full text
A customer searches for “1.4571 cylinder-head m8”. Standard search treats “1.4571” as an order number. Result: zero. In reality it is a material code (V4A stainless steel) that belongs as a filter.
3. Synonyms are missing
“V2A”, “A2”, “1.4301”: three names for the same material. If you search for the wrong one, you find nothing. Manual synonym maintenance works, but does not scale across 280,000 products.
What these 38% mean in concrete terms
We modelled this in an example calculation: a B2B shop with €150,000 daily revenue, 30% of it from search. With 38% failed searches and an average conversion rate of 4%, this means:
- Lost revenue per year: roughly €2.7M
- Realistically recoverable: 60–70%, depending on assortment depth
- Eywora license by comparison: around €16K per year
This is not marketing optimism. This is the order of magnitude in which search problems hit, once you measure them honestly.
What you can do today, without a new vendor
Step 1: Measure zero-result queries
Most shop systems show the most-searched terms with no hits in the backend. If not: a simple logging script is enough to start. That list is your gold mine.
Step 2: Maintain the synonym top 10 manually
Take the top 10 zero-result queries. Check: does the product exist? Under which term? Add a synonym. That alone catches 30–40% of the losses immediately.
Step 3: Measure search share
How much of your revenue actually comes through search? If you don’t know: that is your priority for next week. Measure first, then decide.
When a switch pays off
Standard search systems are enough for small assortments with simple language. Anyone with technical attributes, norms, material codes or multiple languages typically gains +20 to +30% conversion on search traffic with AI-based search. More on this in the direct comparison.
The honest short version: if you have more than 20,000 products with B2B attributes and cannot maintain the search regularly, a switch delivers ROI in 95% of cases. For smaller setups, often only once a specific pain becomes concrete.