Site Search Analytics: Your Zero-Result Searches Are a Shopping List

Your search box is the one place shoppers tell you, in their own words, what they want. Here is what WooCommerce, Magento and GA4 actually record about searches, the blind spots in each, how to log result counts, and a way to sort zero-result terms into fixes and buying decisions.

Most store data tells you what people bought. The search log tells you what they were looking for, typed in their own words, including the things you do not sell and the things you sell under a different name. It is one of the cheapest demand signals a store has, and one of the least used.

The failure point is well documented. Baymard Institute’s benchmarking found that nearly 50 percent of sites fail to give users an effective way to recover from a search that returns no results. Fixing that page helps. Reading the searches that land on it helps more, because they tell you what to fix in the catalogue, in product data and sometimes in what you stock. This guide covers site search analytics for WooCommerce and Magento stores: what gets recorded, what does not, and a weekly routine for turning the log into action.

What your platform records about searches

The three common sources each see a different slice, and none of them sees everything.

SourceWhat it recordsWhat it misses
GA4 enhanced measurementA view_search_results event with the search_term, triggered when the URL contains q, s, search, query or keywordHow many results came back; visitors who declined consent or block the tag
Magento Search Terms reportEvery query with the number of results and the number of uses, under Reports, Marketing, Search TermsWhat the shopper did next; which searches led to orders
WooCommerce coreNothing; searches are not loggedEverything, unless a search plugin or your own code keeps a log

Google’s documentation lists the five default parameters and notes that the search_term parameter populates the Search term dimension. WooCommerce product searches use ?s=, so they are picked up without configuration. Magento’s report is more useful than most people realise: Adobe’s documentation notes that the number of results is updated every time a shopper executes a search, which means a term that returned zero results last week shows zero until someone searches it again after you fix it.

The two blind spots in GA4

GA4 knows a search happened and what was typed. It does not know whether the search found anything, so a zero-result search and a search with 200 results look identical in the report. The second gap is specific to WooCommerce: when a search matches exactly one product, WooCommerce skips the results page and sends the shopper straight to the product. The behaviour is controlled by the woocommerce_redirect_single_search_result filter, which is on by default. No results page means no view_search_results event, so your most successful searches never appear in GA4 at all.

What GA4 sees for a WooCommerce search, by number of results A search with zero results shows a no results page, and GA4 records the term but not that it found nothing. A search with exactly one result redirects to the product page, and GA4 records no search. A search with many results shows the results page, and GA4 records the term. Shopper searches /?s=…&post_type=product 0 results: “No products found” page 1 result: redirect to the product page 2 or more results: results page Term recorded, “zero” is not No search event at all Term recorded correctly
Default WooCommerce and GA4 behaviour. Only the bottom path is recorded fully, which skews any search report toward broad queries.

Log the result count yourself

Both gaps close with a few lines in a child theme or a small plugin. The snippet below turns off the single-result redirect and pushes the search term and the number of matching products to the data layer on every search page:

// Show a results page even when only one product matches
add_filter( 'woocommerce_redirect_single_search_result', '__return_false' );

// Push the term and the result count to the data layer
add_action( 'wp_footer', function () {
    if ( ! is_search() ) {
        return;
    }
    global $wp_query;
    $term  = get_search_query();
    $count = (int) $wp_query->found_posts;
    ?>
    <script>
      window.dataLayer = window.dataLayer || [];
      window.dataLayer.push({
        event: 'site_search',
        search_term: <?php echo wp_json_encode( $term ); ?>,
        search_results: <?php echo $count; ?>
      });
    </script>
    <?php
} );

In Google Tag Manager, send that as a GA4 view_search_results event with a search_results parameter, register the parameter as a custom dimension, and turn off site search in enhanced measurement so the event is not counted twice. Two cautions. Shoppers sometimes type email addresses or order numbers into the search box, and Google does not allow personal information in GA4, so strip anything that looks like an email or a long number before pushing it. And if you keep the redirect for user experience reasons, push the event from the product page instead when the referrer is a search.

If you would rather not depend on GA4 at all, write each search to a database table (term, result count, timestamp, session ID) from the same hook. A server-side log is not affected by consent banners or ad blockers, and it is the only way to see every search.

Six numbers worth tracking

MetricDefinitionWhat a bad value points to
Search usage rateSessions with a search divided by all sessionsA hidden or slow search box, especially on mobile
Zero-result rateSearches with no results divided by all searchesMissing synonyms, typos, unindexed attributes
Refinement rateSearches followed by another search in the same sessionResults that do not match intent, even when not empty
Search exit rateSessions that leave from a results pagePoor ranking or a dead-end no-results page
Search conversion rateSessions with a search that end in an orderCompare with sessions without search; it should be higher
Revenue per searchRevenue from searching sessions divided by searchesThe figure to watch after any search change

Mobile deserves its own view. Metorik’s 2026 WooCommerce benchmark found that 72 percent of orders are now placed on mobile. On a small screen, deep category menus are slow to use, so the search box tends to carry more of the navigation, and a broken search costs more there.

Sort zero-result terms into four groups

A raw list of failed searches is hard to act on. Two questions sort it: do you sell what the shopper asked for, and could search find it?

Sorting zero-result searches by whether you sell the item and how often it is searched A two by two grid. You sell it, searched often: fix search first. You sell it, searched rarely: fix in batches. You do not sell it, searched often: a buying decision. You do not sell it, searched rarely: point to related products or content. How often it is searched Often Rarely Do you sell it? Yes No Fix search first Synonyms, typos, attributes. Shoppers want something you have and cannot find it. Fix in batches Add to the synonym list during the monthly review. Buying decision Repeated demand for a product or variant you do not carry. Send to merchandising. Redirect Point to the closest category or a help page, and move on.
Top left is lost revenue on stock you already hold. Bottom left is free market research.

Most zero-result searches fall into a handful of causes, and each has an owner:

CauseExampleFixOwner
Vocabulary mismatch“hoodie” when the catalogue says “sweatshirt”Synonym list or extra keywords in product dataMerchandising
Misspelling“bluetoth speaker”Typo tolerance in the search engineDeveloper or search plugin
Attribute not searchable“waterproof jacket”, “size 12”Index attributes and descriptions, not just titlesDeveloper
SKU or model number“XR-450”Index SKUs; WooCommerce core search does not include them by defaultDeveloper
Product you do not stock“usb c charger” in a phone case storeWeigh demand; stock it or redirectBuying
Not a product“returns”, “tracking”, “gift card”Redirect to the right help pageCustomer service

Magento has a built-in shortcut for the last two: under Marketing, SEO and Search, Search Terms, you can set a redirect URL for a query, so a search for “returns” goes straight to the returns policy.

Rank the list by what it might be worth

To decide what to fix first, estimate the revenue each failed term might be leaving behind: monthly searches, multiplied by your search conversion rate, multiplied by average order value for that category. It is a rough upper bound, not a forecast, because some of those shoppers would not have bought even with perfect results. Label it as an estimate. Its job is ordering the list, not predicting income.

Zero-result termSearches per monthGroupRough monthly value
hoodie240Fix search first240 x 4% x $62 = $595
waterproof130Fix search first130 x 4% x $88 = $458
usb c charger95Buying decision95 x 4% x $25 = $95, if stocked
returns310RedirectNot revenue; a support cost
Illustrative figures with a 4 percent search conversion rate. Use your own rate and category order values.

Then fix the page itself

Some searches will always come back empty. What the shopper sees next decides whether they stay. Baymard recommends a no-results page that offers related categories, alternative searches with a preview of the top products, recommendations based on browsing history, contact options, and popular products as a last resort, and notes that search tips on their own are not enough. Once your result counts are logged, you can measure the change: the exit rate from zero-result pages should fall after the redesign.

A 30-minute weekly routine

  1. Export last week’s zero-result terms with their counts (GA4 with the new dimension, Magento’s Search Terms report, or your own log).
  2. Normalise them: lower case, trim spaces, merge plurals.
  3. Sort each new term into one of the four groups and assign the owner.
  4. Check last week’s fixes: the same terms should now return results, and their exit rate should drop.
  5. Once a month, send the “buying decision” list with counts to whoever plans the range.

Watching the zero-result rate over time is also a cheap health check. A sudden jump after a theme update or a catalogue import usually means search has stopped indexing something. QuantumLayers explains how to separate that kind of break from normal variation in Anomaly Detection in Data Analytics.

Sources