Shopify search synonyms should help shoppers find the product they mean without erasing important differences between products. In a beauty catalog, a broad rule can mix formats, shades or concentrations that the shopper deliberately specified. I would review a small set of real queries and test both the intended matches and the products that must stay out.
1. Check the search engine before changing a rule
In SwiftSearch, I put a Create synonym action beside zero-result queries, ordered by frequency. The same admin view reports clicks and add-to-carts for popular searches. That connects a proposed rule with evidence that a shopper struggled, instead of beginning with a generic dictionary.
This article develops a review process around that documented feature. It does not assume that every Shopify theme or third-party search app uses the same synonym engine. Identify what serves the predictive panel and the full results page before choosing where to make a change.
Shopify's search behavior documentation describes native matching and typo handling. Test the current result before creating a rule for a misspelling. A synonym list should not become a second implementation of behavior the active engine already handles.
Also distinguish onsite search from Google search. This worksheet concerns what happens after someone uses a store's search box. A successful synonym test is not evidence of a higher Google position; external discoverability is the subject of my large-catalog SEO guide.
2. Sort failed queries into different jobs
For each candidate query, write down what a satisfactory result would contain. Then find an actual eligible product that meets that description. If you cannot identify one, ask whether the right action is a buying decision or a helpful empty state.
I would split the review into vocabulary, availability, catalog data and relevance. A vocabulary problem can justify a synonym. A product with the wrong attributes needs a data correction, while a matching item buried below unrelated products needs a ranking investigation.
| Query relationship | Candidate action | What to verify |
|---|---|---|
EDP and eau de parfum | Review as an abbreviation pair | Intended concentration remains explicit |
face wash and facial cleanser | Review against the actual assortment | Exclude products that do not meet the shopper's task |
rose perfume and rose body wash | Keep the format distinction | A shared scent word is insufficient |
| Two nearby foundation shade codes | Preserve the exact shade | Similar codes are not interchangeable |
| A requested brand the store does not carry | Record an assortment gap | Do not quietly substitute another brand |
These are proposed test cases, not production search logs or a dictionary to import. A merchant still needs to check product descriptions and the language customers use. The purpose of the table is to make a decision inspectable, including a decision to add no rule.
3. Give every candidate an exclusion test
Suppose an illustrative store wants EDP to find products described as eau de parfum. I would record a known matching product and a product with a different concentration. The rule must help the first query without turning a deliberately specific search into a generic fragrance listing.
Test the abbreviation alone, the full phrase, and both within a longer query that contains a brand or size. Also search for the neighboring term that should remain distinct. A rule that looks harmless in isolation may broaden a longer query in an unexpected way.
For a search system with bidirectional groups, test every term as the input. For a directional mapping, document its direction explicitly. Verify the behavior your engine actually implements instead of assuming that a saved group label describes it completely.
My review record would contain the query, intended meaning, matching product IDs, excluded product IDs, search surface and reason for the change. It would also keep the previous configuration. This turns rollback into a specific edit rather than a search for what used to work.
4. Test the panel and the full results page
Shopify documents differences between search customization and predictive search. For example, native semantic search does not apply to predictive search. Do not assume two surfaces produce identical results just because they share an input box.
Run each candidate through the actual mobile panel and the full results page. Record the selected language and market, and keep those conditions consistent during the comparison. If an app owns one surface and the theme owns another, give the discrepancy to the right system owner.
Try a fresh session as well as a repeated search. Check exact product and shade names, not only the broad query that motivated the new rule. The existing search diagnosis overview helps separate vocabulary problems from slow loading and confusing result cards.
Search settings also interact with product availability. Shopify's customization guidance includes controls for unavailable products, so record the active policy before deciding that a missing product indicates a synonym failure. A rule should not be judged using a product that the search configuration intentionally excludes.
5. Measure whether the results became useful
Zero results becoming ten results is a count change, not proof of a better shopping experience. I would inspect whether relevant products are visible and whether shoppers proceed to product views or cart actions. Compare the same query group and note stock or assortment changes during the period.
SwiftSearch's case study describes line-item attribution and consent-dependent tracking. That is why I keep the measurement denominator explicit: tracked interactions are not all store visits. I would not convert a change in a small tracked sample into a claim about total revenue.
Keep a short regression set beside the rule list: known brand searches, exact product names, important shades, format-specific terms and intentionally empty queries. Re-run it after catalog imports or relevance changes. The list should reflect your assortment rather than an arbitrary target number of synonyms.
Start with the most frequent vocabulary gap for which you can name a real matching product. Save the before-and-after results and the exclusion test. If the correct rule remains unclear, book a search review with the query, expected product and current search provider.