The Resolution Layer · Case Study in BriefEKOM
Refrigerators.
A cold read of the complete refrigerator category of a major-appliance retailer — where a fridge is chosen by whether it fits the opening, holds the household's food, and runs on the outlet already in the wall. One of a set of anonymized EKOM analyses — what surfaced, and where it leads.
Lead finding · one electrical label holding two different numbers
167 of 258 published wattages describe the wall circuit, not the refrigerator.
They cluster on 1,650, 1,725 and 1,800 watts — 15 amps at 110, 115 and 120 volts, the household circuit manufacturers state as a supply requirement. On one compact fridge the page reads 1 amp and 1,725 watts in adjacent rows, fifteen times apart. It is not alone: two refrigerators have voltage and amperage transposed, and 24 publish cords measured in inches under a label that says feet.
Also surfaced
Capacities that cannot add up
A freezer larger than the refrigerator it sits in; fresh food of one cubic foot on a full-size model; freezerless units with a freezer capacity. Ten of eleven findings critical, each failing arithmetic on numbers already on the page.
Copy written for another model
Ten refrigerators open their overview as a 23 cu. ft. side-by-side, whatever they are. A stainless unit is written up in a glass finish; a model with no dispenser is described with one.
The layer machines read
On four in five refrigerators, the structured description is escaped page markup; at least 48 carry a one-line placeholder; structured dimensions run about four inches larger than the specification table.
Filled in, and wrong
31,068 cells hold a literal dash that means "none" in some fields and "unknown" in others. A completeness check scores every one of them — and every wrong wattage — at a hundred percent.
How EKOM read it. The engine read every refrigerator page cold — no file, no credentials, no schema — and captured both the specification table and the structured data behind it. It profiled what each field is for from how the category itself uses it, then held every model to the page's own arithmetic and to its siblings: volts times amps against watts, fresh food plus freezer against total, one finish against its twin. Headline values were re-read on the live pages, and one finding that did not survive that check was withdrawn rather than reported.
From diagnosis to channel-ready. Most of the 127 findings collapse into six upstream mechanisms — a circuit value written in as a default, values landing in the neighbouring field or unit, copy assigned by family rather than model, a machine layer built from a different source, every finish keyed as its own record, and a schema with no place for some answers. Fixing the findings fixes the findings; fixing the mechanisms fixes every model that has passed through them and every one that will. Then the same intelligence holds the line at intake, so the next circuit rating in a wattage field never reaches a live page.