The Resolution Layer · Case Study in BriefEKOM
Heating, cooling & water-heating equipment.
An automated read of the public product pages of an equipment manufacturer that sells through distributors and contractors — where the model number is the most reliable field on the page, and a buyer chooses by fuel, capacity and dimensions rather than looks. One of a set of anonymized EKOM analyses — what surfaced, and where it leads.
Lead finding · fuel type that contradicts the model
Propane water heaters whose fuel field says natural gas.
On a line of commercial gas water heaters, the propane models carry the propane designation in the model number and the page address. Of the propane models re-read live, not one lists propane alone: six say natural gas, and thirty-nine list both gases. Every natural-gas model re-read lists natural gas only. A contractor filtering for propane misses the six that say natural gas; one filtering for natural gas finds all forty-five. Fuel type decides the valve, the orifice and the supply line, so this is an error found at start-up, not at the order.
Also surfaced
A heading that names another model
On 213 pages across seven model families, the page heading carries a neighbouring family's name while the title and SKU name the model itself. Anything that takes the heading as the product name records the wrong model. Most of the seven heading findings trace to this one pattern — EKOM's reading, for the manufacturer to confirm.
Specifications that cannot be right
A commercial packaged unit five inches tall. A 65-gallon heat pump water heater published with a 50-gallon tank. Tonnage and capacity from the next size down, and a coil carrying a system's efficiency rating.
Placeholders printed as values
A stand-in number in a tankless heater's hot-water delivery fields, "NULL" as an install location, an editor's "(Duplicate)" note in a model name. Every one passes a completeness check.
What a machine is told
On the page checked live, the visible breadcrumb runs through the category path; the machine-readable one is two steps ending in "product". The sitemap lists dead products and doubled entries. Product identity is structured on effectively every page — which is why those fields have to be exact.
How EKOM read it. The engine read every product page in the manufacturer's own sitemap cold — no file, no credentials, no schema — taking each reported value from its named specification label rather than from the page copy. An early candidate built from page text picked up a cross-sell panel's figures, and was withdrawn. It held every field against the model number and against the sibling that gets it right, then re-read the headline findings live. Four the live re-read did not show — corrected since, or never as first reported — were held back rather than reported.
From diagnosis to channel-ready. Most of the 44 findings collapse into five upstream mechanisms — model pages carrying the family's text, a fuel value set for the line rather than the unit, values taken from the neighbouring size, database defaults exported as content, and a publishing layer generated apart from the product set. Fixing the findings fixes the findings; fixing the mechanisms addresses the models already filed through them and the ones that follow. Then the same intelligence holds the line at the source, so the next model filed under the wrong fuel is caught before a contractor quotes from it.