Ecommerce Site Search: Fix Results That Hide Your Best Products
What ecommerce site search should show a Shopify shopper
Ecommerce site search should return in-stock products that match the shopper's words and that already convert when people reach them. A useful result set is relevant first, available second, and ranked so the strongest SKUs are visible without a scroll.
Search is a different surface from a collection. The shopper typed an intent. If the first screen shows a sold-out variant, a loosely related accessory, or nothing at all, that shopper often leaves instead of browsing the rest of the catalog. Baymard's large-scale ecommerce UX research has repeatedly found that on-site search is used by a large share of visitors and that poor search is a common reason people abandon a site. Treat those findings as a reason to audit your own queries, not as a promise that a better search box will raise your conversion rate by a fixed amount.
A search results page is still a product listing page. Position, filters, and sold-out tiles can hide a winner the same way they do on a collection. The definition of that surface, and how to tell a filter problem from a ranking problem, lives in what a product listing page is on Shopify. This guide starts one step earlier: the actual query the shopper typed.
What you need before you change anything
You need three inputs, and they do not all come from the same place.
- A query report: the phrases shoppers searched, how often, and whether the search returned products. Shopify's Search & Discovery app can store this. Other sources are a search app's analytics, or a storefront event you already log.
- Product evidence: sessions or views, orders, and conversion rate per product, plus whether the SKU is available to sell.
- A short list of products you already know convert and have stock. You will use this list to spot results that bury them.
Skymetrics can rank products by traffic, conversion, stock, and collection exposure so you know which SKUs search should not hide. It does not replace a query report. Pull the phrases from Shopify or from whatever tool actually records searches.
Work from a recent window with enough searches to act on, often the last 28 to 90 days. A phrase searched twice is a note. A phrase searched every week is a fix.
Find zero-result and misleading queries
Open the native search report
In the Shopify admin, open Apps, then Shopify Search & Discovery, then Search, then Search analytics. Shopify's help documentation says this report lists the top online store searches, including searches that returned no results, and that it can be filtered by the last 30, 60, or 90 days. The app must be installed, and search data is collected only after it is installed, so an empty report can mean "not tracking yet" rather than "nobody searches."
If you use another search app, use that app's query report instead. Do not mix two ranking systems in the same audit. Note which engine produced the results you are looking at.
Sort queries into three piles
Export or copy the top queries and label each one before you edit products.
- Zero results. The shopper got an empty state. These are the first fixes if the product exists under a different name.
- Misleading results. Products came back, but the first screen does not match the words. A search for "wool beanie" that leads with cotton hats is this pile.
- Relevant but poorly ordered. The right products exist in the results, yet sold-out, low-converting, or loosely related items sit above available winners.
Also scan for near-duplicates: "t shirt", "t-shirt", "tee", and a misspelling of your bestseller's name. Group them. One synonym or one product title often clears several rows.
Ignore one-off junk queries (random characters, internal staff tests, other brands you do not sell) unless they repeat. A repeating competitor brand query is a merchandising decision, not a typo: either you stock an equivalent and should say so, or you do not and an empty result is honest.
Check whether prominent results are in stock and convert
For each high-volume query, open the storefront search in a private window and record the first 8 to 12 products, in order. That is the screen most shoppers judge. Then score each tile with store data, not with how the photo looks.
- Availability: is at least one variant available to sell at the locations you fulfill online? A product that is published but sold out still occupies a slot.
- Conversion: among products with enough sessions to judge, which ones actually get orders? Use product conversion rate, not storewide conversion rate. The formula and sample-size rules are in the product conversion rate guide.
- Traffic quality: a product with lots of search clicks and almost no orders is a different problem from a product nobody sees.
- Exposure elsewhere: if the same SKU already converts on a collection, burying it in search is a placement miss, not a product miss.
Placement across search, collections, and homepage slots is its own decision. Use ecommerce merchandising decisions when the question is which surface should feature the SKU. Stay in this guide when the question is whether this query's result list is wrong.
A practical priority order for one query:
- In stock and already converting, but below the fold or missing. Move it up or make it match the query.
- In stock, relevant, and unproven. Leave it visible, but do not pin it above a proven SKU.
- Out of stock or unavailable online. Demote it or hide it from search if your theme and search settings allow it.
- Irrelevant. Remove the match by fixing titles, tags, or synonyms, not by hoping shoppers scroll past it.
Do not treat a handful of sessions as proof. If a product has too few views to trust its conversion rate, say so and rank it on relevance and stock until the sample grows. Inventing a "winner" from three views will push the wrong tile to the top.
Fix naming, synonyms, filters, and result placement
Change the smallest thing that makes the query return the right available products. Record the date, the query, and the exact change so the before-and-after comparison later is fair.
Product titles and search terms
Shopify's storefront search matches product data such as title, product type, vendor, tags, and variant options, depending on your search configuration. If shoppers type "navy crewneck" and the product is titled "Indigo Pullover," the engine is not failing. The catalog language and the shopper language disagree.
Put the words shoppers use in the product title when those words are accurate. Use tags or a dedicated search-term field only for words that should match but should not clutter the title, such as a common misspelling or an old product name. Do not stuff every synonym into the title. That hurts the product page and can create false matches on other queries.
Synonyms and redirects
Shopify Search & Discovery supports search synonyms and search redirects. Use a synonym group when two phrases should return the same products ("jumper" and "sweater", "sneakers" and "trainers"). Use a redirect when the query should leave search entirely, such as "returns", "size guide", or "track order".
Add synonyms from the query report, not from a brainstorm. One group per real mismatch is enough. A giant synonym list creates false matches: "tee" matching every product that mentions "steel" is worse than a missing result.
Filters and out-of-stock placement
If the right products exist but shoppers cannot narrow to them, the filter set is the bug. Confirm that color, size, product type, and availability filters on the search page use the same product data as your collections. A filter that works on a collection and fails on search usually means the search template does not render the same filters.
Then decide what happens to unavailable products. Many themes still show sold-out items in the default order. If your highest-converting SKU is in stock and a sold-out colorway ranks above it, demote unavailable products for that results page. Hiding them completely is reasonable when you cannot fulfill the variant and you have an in-stock alternative. Keeping them visible is reasonable when a restock is days away and the product page says so. Pick one rule and apply it consistently.
Pinned or boosted results
Some search setups let you pin products to a query or boost products that match a rule. Pin only when the data supports it: the product matches the words, it is available, and it converts better than the products currently above it. A pin is a manual override. Review pins when stock hits zero or when conversion drops, or the override becomes the new problem.
Boosting every product in a collection is not a search fix. If the catalog itself is too wide or the wrong products are published, fix assortment first. Assortment planning covers which products should stay in the catalog and which should come off the shelf.
Compare search journeys before and after each change
Change one query group at a time. If you rename ten products, add synonyms, and install a new app in the same week, you will not know which change helped.
Before you edit, save a snapshot:
- The query text and its search count in the report window.
- Whether it returned zero results.
- The ordered list of the first 8 to 12 products, with available units and a note on conversion if the sample is large enough.
- The date.
After the change, wait until the same query has been searched enough times to compare, or at least through one normal traffic week. Then check three things only.
- Does the query still return zero results? That should drop to zero for phrases you can honestly match.
- Are the first results available and relevant to the words? Re-open the storefront and compare the ordered list to the snapshot.
- Did product conversion or orders on the newly visible SKUs move in the right direction, without a matching drop on the products you demoted?
Search analytics in Search & Discovery shows top searches and no-result searches. It does not, by itself, prove that a synonym increased revenue. Pair the query report with product-level orders and availability. If orders on the promoted SKU do not move and the demoted SKU loses orders, undo the pin.
A clean operating target, not a guaranteed outcome: zero-result queries that match products you sell should trend down, and the first screen for your highest-volume queries should be dominated by available products that already convert. If that target is missed, the next step is another query fix or a template bug, not a bigger ad budget.
Know when a search-software change is necessary
Fix catalog language, synonyms, filters, and ranking before you replace the search engine. A new app cannot invent a product title shoppers recognize, and it will still surface sold-out products if you tell it to.
A software change is worth a trial when all of the following are true after the audit:
- Query volume is high enough that ranking quality is a weekly job, not a monthly cleanup.
- Native synonyms, redirects, and merchandising rules cannot express the matches you need (heavy typo tolerance, multi-language catalogs, or SKU-level rules your current engine ignores).
- You have confirmed the failure on the live storefront, not only in an app's marketing comparison.
- You can measure the same queries before and after the switch, using the snapshot method above.
Stay on the current engine when the top failures are empty results for products you already sell under different words, sold-out tiles in the first positions, or filters missing from the search template. Those are merchandising and theme fixes. They remain your job after any app install.
If you do trial another engine, run it against the same query list for a defined window, keep the old snapshot, and revert if zero-result recovery or first-screen relevance does not improve. Do not judge the trial by how the search bar looks.
A weekly search merchandising pass
Once the first audit is done, keep the loop short.
- Open the search report and list new or rising zero-result queries.
- Re-check the first screen of your top five queries for sold-out or low-converting products sitting above available winners.
- Apply one synonym, title, filter, or pin change.
- Save the snapshot and review it the following week.
- Hand any SKU that search keeps demanding, and that is running low, to your reorder process. Search demand is a stock signal only if the product is the one shoppers meant.
That pass is the work. The search box is only the place the mistake shows up.