Share of search

The term covers two different measurements. Both are legitimate. They answer different questions, and mixing them produces a number nobody can act on.

Demand-side share of search counts how often people search for your brand versus competing brands. Shelf-side share of search counts how often your SKUs appear when someone searches for a product rather than a brand.

Dimension	Demand-side share of search	Shelf-side share of search
What you count	Query volume for brand names	Result slots your SKUs occupy
Example query	"[your brand] wire stripper" against the same query for each competing brand	"UL listed 600V wire connector"
Data source	Keyword volume tools, Google Trends	Scraped SERPs, marketplace results, site search logs
What it tells you	Brand demand trend	Whether your product data earns placement
Who acts on it	Brand marketing	Merchandising, e-commerce, product data teams

For B2B distributors and manufacturers, shelf-side is usually the more useful of the two. Buyers searching for a 3/8-16 Grade 8 hex bolt rarely type a brand name. They type the spec. Whether you show up depends on whether that spec exists in your data as a structured, matchable attribute.

The rest of this entry focuses on shelf-side measurement, because that is the number product content actually moves.

Shelf-side share of search is a ratio with four boundaries baked into it. Each one has to be fixed before the ratio means anything:

• A category. The measurement is per category, not per catalog. Wire connectors and circuit breakers behave nothing alike, so a blended figure describes neither.
• A query set. Typically 30-100 real queries drawn from site search logs and the terms buyers actually use: spec strings, MPNs, cross-reference part numbers, application phrases.
• A surface and a depth. Amazon results, Google organic, and your own site search are each a separate number. So are top 10 and top 20.
• A count. The numerator is your SKUs appearing; the denominator is total slots captured.

A single category run looks like this:

Query	Slots captured (top 20)	Your SKUs present	Share
3/8-16 grade 8 hex bolt	20	4	20%
3/8-16 x 2 zinc hex cap screw	20	1	5%
grade 8 bolt yellow zinc	20	6	30%
Category total	60	11	18.3%

The measurement only carries meaning as a series. The same query set on the same cadence, monthly for most categories, produces a trend line; changing the set mid-year breaks it unless history is restated. Direction matters more than the absolute number. Hypothetically, a share that rose from 11% to 18% over two quarters describes a working program, while one that fell from 26% to 18% describes a leak.

The queries scoring zero are the diagnostic rows. A zero on "3/8-16 x 2 zinc hex cap screw" for an item you stock usually means the thread pitch, length, or finish is missing or unstructured, not that the SKU is uncompetitive.

The same ratio now applies to LLM answers, and the unit changes. There are no twenty slots. There is one answer, and either you are named in it or you are not.

Surface	What you count	How to capture it
Google organic	Result positions	Scraped SERP
Marketplace search	Placements in top N	Scraped results page
AI Overviews and chat assistants	Citations and named mentions	Rerun the query set, log which domains and products get named

Measure it by running your existing category query set through the assistants your buyers use, several times per query, since answers vary run to run. Record mention rate rather than position, and report it as its own number. Do not blend it into shelf-side share; the denominators are not comparable.

What gets cited skews toward pages that state facts plainly and structurally: a voltage rating in a field, a GTIN that resolves, a spec table instead of a paragraph of marketing copy. That is the same substrate that wins faceted search.

Share of search is an output. You cannot edit it directly. The inputs are unglamorous and they live in your product data:

• Attribute fill rate on the fields buyers filter by. Not overall completeness, but the specific fields that appear in the category's facets. Thread pitch, grade, finish, length. A SKU missing finish is invisible to a finish filter no matter how good the copy is.
• Correct category placement. A wire connector filed under "electrical accessories" is competing in the wrong race.
• Clean identifiers. GTIN and MPN drive matching, cross-referencing, and marketplace eligibility.
• Values from a controlled vocabulary. "Zinc," "Zinc Plated," and "ZN PLT" are three facet buckets holding one product.
• Structured markup on the PDP. Machine-readable specs are what answer engines lift.

This is where the PIM/completion split gets practical. Your PIM stores the finish field and enforces its rules. It does not go find the finish for tens of thousands of legacy SKUs. Anglera does the work of finding and filling it, and writes the values back to the PIM.

Use attribute fill rate on facet-critical fields as your leading indicator and share of search as your lagging one. When the first moves and the second does not, the problem is placement or taxonomy, not completeness.