Contents8
- 01What the search bar tells you about your buyers
- 02Where search fails most often
- 03Searches with no results: the most useful list in the store
- 04Search depends on product data
- 05A page with no results must not be a dead end
- 06Which tool, and when to change it?
- 07Measuring without fooling yourself
- 08Frequently asked questions
A visitor browsing collections is still hesitating. One who types “rabbit 30 cm” into the search bar knows what they want, and tells you word for word. If the page that appears is empty, or full of rabbits of every size except the right one, they do not complain: they close the tab. The lost sale leaves no trace in the dashboard.
This guide is for teams running an online store who suspect their site search is not doing its job. It does not compare search engines: it shows how to read what customers search for, why search fails, and what can be fixed before changing tools.
In short
The words typed into the search bar are the most direct data you have on demand: read them every week.
Most failures come from a vocabulary gap or a feature missing from product pages, not from the engine.
A search with no results is fixed by a synonym, a completed attribute or a product to add, and the empty results page must offer a next step.
The tool comes last, chosen by catalogue size and languages, starting with what your platform already does.
What the search bar tells you about your buyers
The search bar is the only place in the store where customers write their request in their own words. Those words do three jobs: they reveal products people look for without finding them, the real vocabulary of your buyers, and products you do not sell yet.
You still have to record them. Store platforms often have their own search report; failing that, Google Analytics 4 captures them with enhanced measurement, which records a view_search_results event and the typed term in the search_term parameter, provided the results page address carries the query in a recognised parameter such as q, s or search, according to the Google Analytics help on enhanced measurement. A five-minute check: run a search on your store, then see whether it shows up in the report the next day.
Where search fails most often
The Baymard Institute, an independent research institute specialising in ecommerce usability, has been benchmarking the search of leading US and European retail sites for years. Its breakdown by query type shows where engines fall short:
Share of sites that handle each type of search poorly
The slope speaks for itself. An engine almost always finds a product’s exact name, because that name is written on the page. It stumbles as soon as the buyer describes what they want instead of naming it: “soft”, “30 cm”, “baby shower gift”, “machine washable”. According to the same institute, 56% of sites fail to adequately support their users’ search needs.
Key takeaway
The engine only finds what is written
A feature missing from the product page, or written differently from how customers type it, cannot be found, whatever the engine. Most fixes therefore happen in the catalogue.
Searches with no results: the most useful list in the store
Once a week, pull the list of searches that returned nothing, or were followed by an immediate exit. Each line falls into one of four boxes, and each box has its fix:
From an empty search to its fix
- An unknown word
The customer says “cuddly toy”, the catalogue says “plush”. Add a synonym.
- A typo
The engine should tolerate it; if not, a synonym catches it.
- A missing feature
Size or material is not on the product page. Complete the attribute.
- A missing product
You do not sell it. A buying lead, or a page to redirect elsewhere.
The third box is the most expensive to fill and the most profitable to handle, because an added feature serves search, filters, the product page and the feed sent to Google all at once. Our article on the product page that converts covers what buyers look for before paying, and our guide to preparing your store for Google Shopping what Google expects from the same data.
Non-product searches, about shipping, returns or payment, deserve an answer too: a well-titled help page attached to those words saves a customer service question. The returns policy in particular should be easy to find, as our article on the hidden cost of returns points out.
Search depends on product data
A search engine reads the catalogue’s fields: title, description, type, attributes, tags. If size lives in a sentence of the description for some products and nowhere for others, neither search nor a “size” filter can be reliable.
From the field
Filters come from coverage, not intuition
On an online plush toy store we are putting back on its feet, we reviewed 653 products and defined 26 fields before extracting anything; 481 product pages were enriched, all reviewed. Size, the first filter everyone thinks of, was readable for only 41% of products. The order of filters now comes from that coverage report, and a collection page is only created from five products upwards.
The large-scale version of this attribute work, when the catalogue counts thousands of references, is the subject of our guide to enriching a large catalogue with AI under control. For a store in several languages, the question arises in each of them: a French synonym is useless to a German buyer, and it is one of the tasks of the multilingual, multicurrency store.
A page with no results must not be a dead end
Even well tuned, search will sometimes come up empty. The page shown then decides whether the visitor stays:
- Restate the request: show what the engine understood, and suggest the closest spelling.
- Offer a next step: related collections, best sellers, a link to contact the store for a specific request.
- Show unavailability rather than hiding it: an out-of-stock product shown with its restock date beats an empty page, provided the displayed stock is right, which is what our article on inventory sync covers.
One technical precaution: search results pages are not meant to be indexed. If every search creates an address that search engines can crawl, the store fills up with pages without content of their own, a trap described in our article on forms and search on a static site.
Which tool, and when to change it?
The tool comes last. Before considering one, look at what your platform already does: on Shopify, the Search & Discovery app, published by Shopify itself and offered at no extra charge according to its App Store listing, handles synonyms, filters, product boosts and a search report.
The platform’s search, well tuned
No extra component to maintain or pay for
Synonyms and filters set by hand, every week
Enough as long as the catalogue stays readable and the language single
A dedicated search engine
Finer typo tolerance and ranking
Useful when the catalogue counts thousands of references or several languages
A subscription, one more script on the pages, a dependency
That extra script is not a detail: every feature added weighs on speed, and our article on store speed shows what that speed is worth in sales. Before installing anything, the calculation in our article on the cost of Shopify apps applies, and what is native or not depends on the platform, as our Shopify or Medusa comparison shows.
Measuring without fooling yourself
Three numbers are enough, tracked over time rather than compared with a market average:
- The share of searches with no results, which should fall as synonyms and attributes are added.
- The share of searches followed by an exit, which tells you whether the results matched the request.
- The conversion rate of visits with a search, to compare with the store’s other visits.
These sit naturally next to the five numbers to track every week. The most searched words also feed other work: the ads and keywords of a Google Shopping campaign, the titles of the listings you publish on a marketplace, whose engine follows its own rules (see selling on Amazon alongside your store), or the questions that come back in customer reviews.
A buyer who found the product without buying can still be won back: our guide to abandoned cart reminders and our article on welcome, post-purchase and repeat-purchase emails cover what comes next. And once search works and orders arrive, it remains to know what each one costs to collect, the subject of our article on real payment fees. The other decisions about a store are gathered in our ecommerce section.
If your site search seems to miss what customers ask for, our ecommerce support starts with your data: what customers type, what your product pages contain, and the gap between the two.
Frequently asked questions
How do I find out what customers search for on my store?
Start with your platform’s search report, if it has one. Otherwise, Google Analytics 4 records site searches through enhanced measurement, provided the results page address carries the query in a parameter such as q, s or search. Export the list of terms every week, and isolate those that returned no results: they are the ones to fix first.
Does an online store need a paid search engine?
Not necessarily. As long as the catalogue stays a reasonable size and in a single language, the platform’s search, with well-tuned synonyms and filters, covers the essentials. A dedicated engine makes sense when the catalogue counts thousands of references, several languages or complex queries, and once product data is clean: it will not find a size missing from the product pages.
What should a no-results search page show?
What the engine understood, a spelling suggestion, then a next step: related collections, best sellers and a way to contact the store. The goal is that the visitor always has somewhere to go. Then add the term to the weekly list of empty searches, so that you fix the cause rather than just the page.
Should search results pages be indexed by Google?
As a rule, no. Every search could create one more address, and the store would fill up with pages without content of their own that search engines would crawl instead of your product pages. If a search comes back very often, it is better to create a real collection page, with a title, some text and a chosen selection of products.
Sources
- Google Analytics Help, enhanced measurement events (new tab)support.google.com
- Shopify App Store, Search & Discovery (new tab)apps.shopify.com
Jim · founder of Brain Plus AI
Also in e-commerce
Practical guide5 min
Product page that converts: what buyers look for
Before paying, buyers check four things: what exactly the product is, what they will pay in total, when they will receive it and what happens if it does not suit them. A page that answers those questions sells better than a page full of fine phrases.
Practical guide6 min
Online store: five numbers to track every week
An online store dashboard shows dozens of metrics; five are enough to decide every Monday. Which ones, how to read them together, and what each one prompts you to do.
Practical guide8 min
A catalogue of thousands of products: AI under control
On a hundred product pages, one person reviews everything; on ten thousand, full review becomes impossible, and a faulty prompt repeats itself everywhere overnight. The process that keeps control of mass enrichment: sorting, sampling, batches, automated checks, targeted review and the ability to roll back.