On-site search is only as good as your attributes
Faceted and on-site search run entirely on structured attributes — here's why thin product data quietly kills conversion, and how enrichment fixes it.
Shoppers don't browse anymore, they search and filter. But every facet on a category page — size, material, voltage, connection type — is really just a product attribute wearing a UI. When the attribute is missing, blank, or buried in a PDF, the facet doesn't work, and the product effectively doesn't exist for anyone using it. Here's why that gap is more common than most distributors assume, and what closes it.
A facet is just an attribute with a checkbox
Faceted navigation looks like a search feature. Mechanically, it's a database query against structured fields. When a buyer clicks "3/4 inch" or "IP65" or "linen," the search engine filters on a discrete attribute value — not on a paragraph of marketing copy that happens to mention the spec somewhere in the middle. If that field is null in the underlying catalog, the product is invisible to that filter, full stop. It doesn't rank lower. It doesn't get buried on page three. It disappears from the result set entirely, because the query never matches an empty field (BigCommerce).
This is why "the description mentions it" isn't the same as "the attribute is populated." Search engines, faceted or otherwise, don't parse prose for specs in real time at query volume — they index structured fields. A supplier PDF that says "NEMA 4X rated" in paragraph three is invisible to a NEMA rating facet unless someone (or something) extracts that value into its own field first.
The zero-result problem is a data problem wearing a UX label
Retailers treat zero-result searches as a search-tuning issue — synonyms, stemming, typo tolerance. Sometimes it is. More often, the query is fine and the catalog simply doesn't have the attribute filled in for the products that would otherwise match. Industry benchmarks put zero-result rates in the 10-20% range across ecommerce catalogs (Wizzy), and site search traffic is disproportionately valuable — search visitors convert at meaningfully higher rates than browse-only visitors, with some retailers reporting conversion multiples of 2x to 6x when search actually surfaces the right product (Algolia). Every zero-result session on that traffic is a buyer who typed the exact thing they wanted and got told it doesn't exist.
The abandonment behavior compounds the cost. A large majority of shoppers say they're likely to leave and buy elsewhere after a failed search, and a meaningful share say they'll avoid a site outright after repeated bad search experiences (Algolia). That's not a one-time lost cart. It's a buyer who now defaults to a competitor's site for the next search, and the one after that.
Where the gap actually lives
| Facet a buyer expects | What it requires in the data | What happens when it's blank |
|---|---|---|
| Size / dimension | Normalized numeric field, single unit of measure | Product excluded from every size filter, even if the size is right |
| Material | Controlled vocabulary value (not free text) | "Linen" search misses products described as "flax fabric" |
| Voltage / connection type | Extracted spec, not embedded in a spec-sheet PDF | Facet count shows "(0)" even though matching SKUs exist |
| Certification (UL, NSF, IP rating) | Structured boolean or enum field | Compliance-driven buyers filter it out before they ever see it |
| Compatible accessories | Relationship field between SKUs | Cross-sell facet never fires, upsell opportunity lost |
Each blank cell isn't a cosmetic gap. It's a buyer who searched with intent and got a "no results" page, or worse, saw a shorter list than the catalog actually supports and assumed the retailer just doesn't carry it.
Before and after: what enrichment actually changes
Here's a typical raw feed description for a circulator pump, pulled straight from a supplier catalog:
"High-efficiency circulator pump for hydronic heating systems. Compact design, easy install, energy-saving operation."
That copy is fine for a product detail page. It is useless for faceted search — there isn't a single filterable value in it. Compare that to the same SKU after enrichment extracts and structures the attributes that were already sitting in the supplier's spec sheet:
| Attribute | Value |
|---|---|
| Flow rate | 11 GPM |
| Head pressure | 19 ft |
| Connection size | 1-1/4 in NPT |
| Voltage | 115V |
| Motor type | ECM variable-speed |
| Energy Star listed | Yes |
| Mounting orientation | Horizontal or vertical |
Nothing here is invented. Every value is extracted from the supplier's own documentation and quality-scored against the category's expected schema — the facets just didn't exist as data until someone pulled them out of the PDF and normalized them into fields the search index can actually query.
Ask an answer engine
The same gap shows up outside your own site. Ask an AI shopping assistant "which circulator pumps handle 1-1/4 inch NPT at 115V with Energy Star certification" and it can only surface SKUs where those attributes exist as clean, structured data — the model isn't going to parse a marketing paragraph to guess at a connection size. A catalog that only has "compact design, easy install" in a description field is as invisible to an answer engine as it is to your own faceted search.
The fix is upstream, not on the search index
Retailers often respond to weak facets by tuning the search engine — better synonyms, fuzzy matching, re-ranking. That helps at the margins, but it can't manufacture a value that was never in the catalog. The fix has to happen upstream, in the product data itself: scoring every SKU against what its category should have, gap-filling the missing fields from supplier and source documentation, and keeping that data current as suppliers update specs.
That's the layer Anglera works in. Your PIM — or your flat file, if you don't have one — stores the data; Anglera continuously scores, gap-fills, and enriches it so the attributes your facets and your AI answer engines depend on actually exist, extracted from real source documents rather than left blank or guessed at. It plugs in additively, alongside whatever catalog system you already run, and most catalogs are enrichment-ready within about 30 days. Faceted search was never really the problem. The attributes underneath it were.