The Resolution Layer · Case Study in BriefEKOM
Agricultural equipment.
A cold read of the used and new agricultural machine inventory of a multi-branch equipment dealer — where one listing is one physical machine, and a buyer decides on its hours, its warranty and its price before anything else. One of a set of anonymized EKOM analyses — what surfaced, and where it leads.
Lead finding · machines whose own record cannot be true
Planters reading tens of thousands of engine hours — on machines with no engine.
22 of the dealer's 26 planters carry an engine-hours value, and five read from some 12,500 up to over 40,000. On the headers, 74 of 109 carry engine hours too. The large values read as acre or bale counts stored in an hours field, and the dealer's own classification codes say these are implements. Separately, 315 units carry a warranty that "expired" on 1 January 1975 — the same date on every one, which fits an empty value published as a real date — while the "has warranty" flag reads No on every unit, including 255 whose warranty has not yet expired. The hour count is the number the price is negotiated on, and the warranty is what the buyer is buying; on these machines the record gives neither.
Also surfaced
Titles that contradict the listing
36 units whose page title and listing name disagree: harrows titled as rock pickers and sprayers, a vertical tiller titled as a header trailer, a wheel tractor titled and filed as a track tractor.
Codes that contradict the machine
A rotary cutter carrying the codes for tires; a vertical tiller carrying the code for sprayers. Wherever the codes group the inventory, the machine is counted in the wrong place.
Prices that mean nothing
Three used machines priced at one cent and four used units with no price at all. A buyer comparing them is comparing placeholders.
New machines carrying working hours
17 units flagged new carry more than 1,000 engine hours. Demonstrators and rental returns are a policy question, not a defect — and the report holds it as one for the dealer.
How EKOM read it. The engine read the dealer's own public site search cold — nothing supplied, no integration — category by category. It profiled the inventory's fields, then held every machine against itself: the hours against the equipment type, the warranty date against the warranty flag, the title against the name and the code. Every extent was recomputed twice by independent code paths, not taken from the automated pass's own count.
From diagnosis to channel-ready. Most of the findings point to four possible upstream mechanisms — one hours field for every kind of meter, an empty date published as a date, titles built apart from names and codes, and placeholders reaching the live record. Fixing the findings fixes the findings; fixing the mechanisms fixes every machine that has already passed through them and every one that will. Then the same intelligence holds the line at intake, so the next bale count filed as hours is caught the day it is listed.