Site search and filters are only as useful as the product structure behind them. If category names, attributes, synonyms, and variant values are inconsistent, shoppers see irrelevant results or no results at all. Taxonomy gives the discovery system the context it needs to rank and refine products.
Why taxonomy affects search
Search needs to understand that a shopper’s wording may differ from the catalog wording. A customer may search for “sofa”, while the catalog says “couch”; “water bottle”, while a product says “hydration flask”; or a model number that appears only in a structured identifier. Taxonomy and controlled attributes create useful signals beyond the title.
Read how bad taxonomy affects site search before changing ranking rules. Fixing the source structure usually has a longer-lasting effect than adding individual synonyms.
Build filters around buying decisions
Start with the questions shoppers ask when comparing products in each category. Filters should narrow a meaningful set of products and help a shopper decide what to do next.
- Use category-specific filters instead of exposing every field in the database.
- Use canonical values for colour, size, material, compatibility, and other repeated concepts.
- Keep product type separate from properties such as finish, audience, or use case.
- Do not show empty filters or values that return one obscure product.
- Make variant information usable without creating duplicate or confusing product cards.
Control synonyms and query intent
Keep a synonym dictionary with the query, preferred term, related terms, language, category context, and owner. Some synonyms should be equivalent; others should broaden the result set. “Sneakers” and “trainers” may be interchangeable in one market, while “running shoes” should often remain a more specific intent.
Use query logs to find repeated failed searches, not just obvious spelling mistakes. A high-volume zero-result query is evidence that the catalog, taxonomy, synonym set, or ranking rules need attention.
Use zero-result reports as taxonomy research
Review zero-result queries weekly and group them into product gaps, vocabulary gaps, data gaps, and intent gaps. A product gap means the range may be missing. A vocabulary gap means the product exists but uses different language. A data gap means the relevant attribute or identifier is missing. An intent gap means the query needs a landing page, collection, or merchandising rule.
Test the complete discovery journey
Test search, category navigation, facets, breadcrumbs, product detail pages, and add-to-cart paths together. A result may look correct in search but fail when the category filter removes it, the URL canonicalizes incorrectly, or the product variant lacks the selected attribute.
Track search exit rate, refinement rate, zero-result rate, result click-through rate, add-to-cart rate, and conversion rate by query group. Compare changes against a stable baseline and annotate taxonomy releases.
Connect search improvements to the product model
Use the faceted navigation guide and product attributes guide to review the underlying model. A PIM or catalog governance process should make it easy to add a new controlled value, approve it, map it to a category, and publish it to search without editing dozens of pages by hand.
For a wider catalog review, use the Catalog Health Score to identify structural and content issues that affect discovery.
Example: turning a failing query into a taxonomy fix
Suppose your search log shows “couch” returning few results while “sofa” performs well. Work through it the same way every time:
- Classify the gap. The products exist, so this is a vocabulary gap, not a product gap.
- Fix it at the right layer. Add “couch” as a synonym for the category and attribute value “sofa”, scoped to Furniture. Do not retitle products.
- Check the facet. Confirm the “Type” filter still shows one value, “Sofa”, not both.
- Record the decision. Log the query, the change, the owner and the date so the next editor does not undo it.
- Re-measure. Compare zero-result rate and result click-through for that query group after two weeks.
Filter design by category
Filters should reflect the way a shopper narrows the choice in each category. The table below shows illustrative starting points; confirm them against your own query and click data.
| Category | Primary filters | Secondary filters |
|---|---|---|
| Sofas | Type, seats, width, colour | Material, depth, delivery time |
| Running shoes | Size, gender, surface, brand | Cushioning, drop, colour |
| Laptops | Screen size, memory, storage, processor | Weight, battery, ports |
| Skincare | Product type, concern, skin type | Size, free-from claims |
| Industrial fasteners | Type, thread, length, material | Finish, standard, pack size |
Put the filters people use first, and hide values with very few products. If a facet value returns one product, it is usually a data-quality issue, not a useful option.
Facet governance rules
- Every facet has an owner, a data type, a unit, and a list of allowed values.
- New values go through an approval step, so “Grey”, “Gray” and “Charcoal grey” do not all appear.
- Numeric facets (width, capacity) use ranges, with the unit shown once.
- Filters apply to categories, not globally, and are inherited from the parent category where sensible.
- Retired values are mapped to a replacement, not deleted, so old URLs and saved searches still work.
SEO and crawl considerations for filtered pages
Facet combinations can create thousands of near-duplicate URLs. Decide which filter pages deserve to be indexed (for example “Sofas > Velvet”, which has real search demand) and which should be blocked from indexing or canonicalized to the parent category. Keep the rules in the taxonomy documentation, not in individual templates. The faceted navigation guide covers the trade-offs in more detail.
Metrics worth tracking
| Metric | What it tells you | Typical action |
|---|---|---|
| Zero-result rate | Vocabulary, product or data gaps | Group queries, fix by gap type |
| Search exit rate | Results that do not match intent | Review ranking and category assignment |
| Refinement rate | Whether initial results were close enough | Check filter order and defaults |
| Filter usage by facet | Which facets help | Promote popular ones, retire unused |
| Result click-through | Relevance of top results | Review synonyms and boosting |
Site search and filters FAQ
Should I fix search with synonyms or fix the taxonomy?
Fix the taxonomy and attribute values first. Synonyms are a good last mile, but they cannot rescue products with missing or inconsistent attributes.
How many filters is too many?
There is no fixed number. If most shoppers never use a filter, or it only returns one or two products, remove or hide it.
How often should I review search logs?
Weekly for zero-result queries, monthly for broader trends, and after every taxonomy or attribute release.
Search queries by intent type
Not every query needs the same fix. Grouping the log by intent shows which part of the taxonomy to change.
| Query type | Example | Best response |
|---|---|---|
| Product type | “sofa”, “mascara” | Land on the matching category with filters |
| Attribute-led | “waterproof jacket”, “velvet sofa” | Return products matching the attribute; consider a filtered landing page |
| Identifier | Model number, GTIN | Match against structured identifier fields first |
| Brand | “Nike” | Brand page or filtered category |
| Problem or use case | “gift for gardener” | Curated collection or guide content |
| Misspelling or variant | “moisturizer / moisturiser” | Normalize through synonyms and stemming |
Handling attribute-led queries
When customers combine a product type with an attribute, the search engine needs both to be queryable. If “velvet” only appears in the product title, it will not behave consistently. Store it as a controlled material or fabric value, so search can use it as a filter and a ranking signal. The same data then powers the facet, the filtered landing page and the feed.
Check mobile behavior separately
Filters that feel fine on desktop can be hard to use on a phone. Test the order of facets, how many values are visible before “show more”, whether selected filters are easy to clear, and whether the result count updates. Review search and filter usage by device, since priorities often differ.
A 30-day optimization loop
- Week 1: export search queries, zero-result queries, and high-exit categories.
- Week 2: fix the highest-impact category, attribute, and synonym gaps.
- Week 3: test ranking, filters, breadcrumbs, and mobile behavior.
- Week 4: compare discovery and conversion metrics, document the decision, and repeat.
Search is a living product experience. A clean taxonomy makes every future optimization faster and safer.
Ecommerce site search and product filters organized by structured attributes
