Why a store needs good search
Shoppers who use search already know what they need: instead of browsing the catalogue, they name the product. That is the clearest signal of intent on the site, so a search failure costs more than a merchandising one — the shopper said what they wanted to buy and got an empty page or the wrong product.
The second effect is data: the query log shows which words shoppers use for products and what is missing from the assortment.
What site search is made of
| Component | What it does | Example |
|---|---|---|
| Autosuggest | Suggests queries, categories and products while typing | “sne” → “sneakers men” |
| Typo tolerance | Finds the product despite a misspelling | “sneekers” → “sneakers” |
| Stemming | Reduces word forms to one base | “running”, “runs” → “run” |
| Spelling variants | Treats regional spellings as one word | “colour” = “color” |
| Synonyms | Links different names for one product | “hoodie” = “hooded sweatshirt” |
| Facets | Narrow results by attribute | brand, price, size |
| Ranking | Sets the order of results | in-stock items first |
| Personalization | Uses the shopper’s history | favourite brands, size, price range |
English inflection is light, but exact matching still breaks on plurals, run-together model names (“iphone15promax” vs “iphone 15 pro max”), US/UK spelling (“grey”/“gray”, “jewellery”/“jewelry”) and the wrong keyboard layout: on a German QWERTZ keyboard Y and Z swap places, so “yoga mat” arrives as “zoga mat”.
How site search has evolved
- Lexical. Looks for the query words in titles and descriptions and ranks with formulas such as BM25. Precise on SKUs and brands, lost on “something warm for ice fishing”.
- Semantic. Turns the query and the products into vectors and compares meaning rather than characters.
- Hybrid. Runs both searches and merges the results: precision on SKUs plus understanding of descriptive queries.
- Conversational. An AI assistant clarifies the need with questions and picks products in a dialogue.
How to measure search quality
| Metric | Formula | What it shows |
|---|---|---|
| Search usage rate | sessions with search / all sessions | How much shoppers rely on search |
| Search conversion | orders in sessions with search / sessions with search | Search’s contribution to sales |
| Zero-results rate | queries with no results / all queries | Where search found nothing |
| Results CTR | queries with a click on a result / all queries | How relevant the results are |
Your analytics platform collects this data if the query is in the URL: GA4 recognises the parameters q, s, search, query and keyword.
Where to start a search audit
- Export the 100 most frequent queries and check the results by hand.
- Review zero-result queries and queries with no clicks separately — that is your list of concrete failures.
- Test your top 10 brands with typos, alternative spellings and the wrong keyboard layout.
- Make sure out-of-stock products do not sit at the top of the results.
- Track search metrics separately for desktop and mobile.
- Validate every ranking change with an A/B test, not by eye.