The Dealer Classification Taxonomy

Flooring, tile, cabinet, and countertop products are typically sold through independent local dealers—showrooms where customers can see materials, get advice, and place orders. But when you search for these businesses online, the results include designers, contractors, manufacturers, and unrelated businesses. Distinguishing actual dealers from adjacent businesses is the first challenge in any systematic research.

A dealer website is not always obvious at scale.

A search for “flooring” in a metro area returns retailers, designers, contractors, brand pages, association sites, distributors, and businesses that have nothing to do with the trade but used the word once on an About page.

A yes/no model creates bad data fast. It pulls in too much noise, excludes too many edge cases, and hides uncertainty that should stay visible.

I used four categories, each with explicit inclusion and exclusion rules.

DEALER

A business that retails flooring, tile, cabinet, or countertop products with a physical showroom or operational presence.

The site must show evidence of products carried, services offered, and location information that supports a customer visit. Pure design firms without product retail are excluded. Pure installation services without product retail are excluded.

The category is intentionally narrow. The research is about local product destinations, not the entire trade ecosystem.

ADJACENT

A business adjacent to the trade but not part of the dealer layer.

This includes interior designers who specify materials but do not retail them, general contractors who install but do not sell, manufacturer brand pages, trade associations, and distributors.

These businesses interact with the dealer layer, but they answer a different question than the research is asking.

Including them in the dealer count would inflate the population. Excluding them entirely would hide how much trade activity sits near dealer retail without actually being dealer retail.

The category exists so the structure of the channel stays visible.

NON_DEALER

A business that surfaced during discovery but does not operate in this space.

Cleaning services. Real estate agencies. Unrelated retail. Businesses that happened to match a keyword but are not part of the flooring, tile, cabinet, or countertop trade.

NON_DEALER is the largest single category in the discovery pool because search-based dealer discovery returns substantial noise.

The category matters because the noise should be counted, not silently removed.

UNCERTAIN

UNCERTAIN is a real category, not a failure mode.

Some sites do not provide enough evidence to classify honestly.

This includes single-page placeholders with no business detail, sites that load with errors, sites in a language the classifier cannot parse confidently, parked domains, and domains that appear to be transitioning.

UNCERTAIN sites are excluded from population statistics rather than forced into a category.

Forcing them would introduce false confidence into the data.

Edge cases that drove rule changes

Dealers operating from a residential address with no showroom are classified as DEALER if other evidence supports the classification. Product lines, services described, customer reviews, business hours, and clear local operating signals all matter. The absence of a showroom alone does not exclude them.

Manufacturer-direct-to-consumer brands can look like dealer sites on a surface scan. They show products. They explain categories. They may even offer ordering paths. But they fail the local product destination test. The classifier weights geographic specificity and local presence over product display alone.

Defunct businesses with operational-looking pages create another problem. Some closed businesses still have websites that serve content. These are flagged as UNCERTAIN unless explicit closure language is present. If the site clearly says the business is closed, it becomes NON_DEALER.

Bilingual sites can also distort classification when one language version is empty or filled with stub content. In those cases, the classifier evaluates the populated version and ignores the empty one.

Validation

Validation against a manually reviewed sample produced 87% agreement between the classifier and human judgment.

The 13% disagreement was concentrated in the ADJACENT and UNCERTAIN categories, which is where the judgment is hardest for humans too.

The DEALER and NON_DEALER categories had agreement above 92%.

That matters because the research depends on the population being clean enough to trust.

This taxonomy is what keeps the population statistics honest.

Without explicit rules, the dealer count would either be inflated by adjacent businesses or deflated by overly conservative exclusions.

The categories themselves are the methodology.