Resolution Layer Case Study · Read & Diagnose
Farm & construction equipment.
A catalog read across the new and used machine inventory of a multi-store equipment dealer group — tractors, combines, loaders, implements and attachments, where one listing is one physical machine and a buyer decides on its hours, its condition and its price before anything else.
This is a real EKOM catalog analysis, with the client's identity removed. The client is a multi-store agricultural and construction equipment dealer group selling new and used machines through a public equipment search. EKOM read every machine that search lists, flagged 20 items, and reports 13 as actionable inventory defects. Seven were withdrawn before the report because they described how EKOM's own read had shaped the record rather than anything about the dealer's listings. None is in any count below. Every extent here was recounted by an analyst against the full inventory; the automated findings' own counts ran low and are not used.
What makes an equipment inventory different from a product catalog is that one row is one physical machine. It has its own stock number, its own serial, its own meter reading and its own location, and a buyer comparing two of them is comparing those numbers — not a feature list. The hours decide the price. The condition flag decides which filter a machine appears in. The size field decides whether a contractor searching for an eight-foot blade finds one. When a field on a machine's record contradicts another field on the same machine, there is no sibling to check it against; the record simply says two things.
A product page that is wrong misdescribes a product. An inventory record that is wrong misdescribes a specific machine on a specific lot — and the number it gets wrong is usually the one the price is negotiated on.
Stock numbers, serials, prices, locations, manufacturer names and the dealer's own category vocabulary have been removed or generalized here, because on a dealer inventory a make, a model and a location together identify the dealer. Counts are exact, and the field values quoted are the ones the record carried.
What's inside
- At a glance — the severity split, where the findings concentrate, and how the read was done.
- What a buyer runs into first — a new tillage disk with 57,220 hours, balers whose "hours" read as bale counts, and machines carrying two different hour readings.
- Priced above the former price — six clearance machines whose asking price exceeds the price they were marked down from.
- Machines the filters can miss — widths entered in inches under a label that says feet, and 71 machines with no industry value.
- Where it traces back, and what's next — the five mechanisms most of these findings read as, and the order to fix in.
Nothing was supplied and no internal system was touched. Everything below was read from the dealer group's public equipment search, and every machine named in the findings was re-read live six days after the first read.
13
Actionable findings
(20 flagged)
Where the 13 concentrate
Theme
What it is
Findings
Hours that cannot be true
A new implement at 57,220 hours; balers near 29,000 and 35,000
2
Two hour figures on one machine
The displayed reading and the stored hour value disagree
1
New machines carrying working hours
Machines flagged new with hundreds or over a thousand hours
1
Asking price above the former price
Clearance machines whose markdown reads as a markup
1
Widths in inches, labeled feet
A 96-foot snow blade; an 84-foot grader box
1
Missing from the industry filter
71 machines with no Agriculture or Construction value
5
No model year
13 machines with a blank year field
1
Stock number filed as a serial
14 attachments carrying their own stock number as serial
1
How EKOM read this inventory
1
Read
The dealer group's own public equipment search — no file handoff, no credentials, no integration.
One row is one physical machine: its own stock number, serial, meter and location.
→
2
Profile
Infer what each field is for from how the inventory itself uses it, then hold every machine to that.
Hours live in two fields and size is a number plus a unit label — which is what makes a disagreement detectable.
→
3
Analyze
Surface defects, rate severity, and bind each one to the machines and values behind it.
13 findings, 8 themes, 5 causes.
Machine-surfaced signals from a cold read, recounted by an analyst — not a dealer-verified defect list. Some will prove intentional, which is why each is tied to the values behind it.
What a buyer runs into first.
The lead finding is about hours — on a used machine the number a buyer reads first, and on a new one the number that should not be there at all.
Critical · Hours that cannot be true
A new tillage disk with 57,220 hours — and first in the hours sort.
A current-model-year tillage disk, flagged new, publishes a meter reading of 57220 Hours. No new implement has run 57,220 hours; the value reads as a figure typed into the wrong field. Sorting the dealer group's entire inventory by hours, high to low, on the live re-read puts this disk first — ahead of every tractor and combine.
Field
What the record holds
Verdict
Condition
new — in the flag, the title and the listing address
agree
Model year
the current year
agree
Hours
57220 Hours
cannot be true
High · Hours that cannot be true
Two large square balers reading some 35,000 and 29,000 "hours".
Two recent-year balers publish readings of about 35,000 and 29,000 hours. At a working season of a few hundred hours those are figures for machines several decades old; they are the right size for bale counts, which a baler's monitor keeps on a second meter. Each machine's detail page has a field for exactly that — a secondary meter reading — and on both it reads N/A. Raised at reduced confidence; reported because the analyst's recount and the live re-read both found the values as published.
A buyer who sorts or filters by hours is using the one field that is supposed to be arithmetic. One impossible value at the top of that sort says the sort cannot be trusted, and that doubt does not stay on the disk. A third shape of the same problem is a policy question rather than a defect: 896 machines flagged new carry an hour reading, 14 of them over 1,000 hours. Dealers sometimes class demonstrator and rental units as new; where that line sits is the dealer's call, and the report holds it as one.
Two hour figures, one machine.
Each machine carries its hour count in two fields — one the listing badge displays, one stored as a bare value — and on the machines below they disagree.
Critical · Two hour figures on one machine
The badge says 900 hours. The record's second hour field says 12.
A late-model four-wheel-drive tractor displays 900 Hours on its card and its detail page while its record's bare hour value holds 12. The analyst's recount finds 31 machines in this state, every one of them flagged on order rather than in stock. On a combine the two fields read 531 and 1,075; on a row-crop tractor listed at just over $300,000, 815 against 1,200.
Shape
What the record shows
Machines
Two fields that disagree
The displayed reading and the bare hour value are different numbers, all on machines flagged on order — gaps from a few dozen hours to 900 against 12
31
A reading with no value behind it
A displayed hour reading but no bare hour value at all; a filter built on the bare value cannot see the machine
144
Two copies of a fact that agree are redundancy. Two that disagree are a coin toss the buyer does not know they are making — and on a $300,000 tractor the hour count is the number the price is negotiated on.
Every one of the 31 disagreements sits on a machine that is on its way rather than on the lot, which reads as two values captured at different moments and reconciled on arrival. Which of the two is right on each machine is a question only the dealer's inventory system can answer. If the hours filter reads the bare value — the field built for it — a buyer who asks for machines under 50 hours is shown a 900-hour tractor; that is a check the dealer can close in one look.
Priced above the former price.
A machine in the dealer group's clearance inventory can carry two prices in its record: the asking price and the former price it is marked down from. On six machines the asking price is the higher of the two.
Critical · Asking price above the former price
A tractor asking just over $500,000 whose record says it used to be some $37,000 less.
A four-wheel-drive tractor carries the clearance flag, an asking price just above half a million dollars, and a former price in its record roughly $37,000 below what it is selling for. The analyst's recount finds six machines in this state — tractors, an excavator and a crawler dozer — and every one of them is in the clearance inventory. The analysis reads the pair as the two prices entered in each other's fields; the record cannot confirm that, and the question is held for the dealer.
Observed · What a buyer sees today
The former price is in the record. On the pages re-read, it is not on the screen.
On the six cards re-read live, and on the one detail page opened, the asking price is shown alone; no crossed-out former price is rendered. The exposure is in the record: wherever it is read with both prices — a crossed-out price on a future page template, a listing feed, a marketplace, an assistant comparing offers — a clearance markdown reads as a markup. Whether that is on a screen today depends on where the field is rendered, which the dealer knows better than this read does.
That is the shape of most of this study: the exposure sits in the record before it sits on the page, and the record is what feeds, marketplaces and assistants read.
Machines the filters can miss.
Two themes with one effect: the machine is in the dealer's inventory, and a buyer using the search the way it is meant to be used may not reach it.
High · Widths in inches, labeled feet
A 96-foot snow blade. An 84-foot grader box. A 168-foot pusher.
An eight-foot snow blade publishes Size: 96 Feet — its width entered in inches under a label that says feet. A laser grader box reads Size: 84 Feet; a snow pusher reads Size: 168 Feet. On each of these the model name carries the true width. The analyst's recount finds 33 blades in this state; the wide sizes on draper heads and sprayers, 40 to 120 feet, are real and were excluded. Raised at reduced confidence; reported because the recount and the live re-read confirmed the values. A contractor who filters for an eight-foot blade does not find it, and one who finds it reads a width no blade has.
High · Missing from the industry filter
Self-propelled applicators that are not in Agriculture. Compact loaders that are not in Construction.
The search divides the inventory into Agriculture and Construction, and the field built for that split is blank on 71 machines. The blanks fall into six equipment types — applicators, spreaders, compact loaders, toolbars, harrows and levelers account for 69 of the 71 — rather than scattering across the inventory, which reads as an industry assigned by type with six types missing from the map. If the Agriculture facet reads this field, a farmer who starts there does not see an applicator. Six type-level rules would cover 69 machines.
Also surfaced: 13 machines with no model year — four of them tires and a bucket that carry none by nature, and others that do — and 14 fabricated attachments from one maker whose serial-number field holds their own stock number, so a serial search matches an inventory tag.
Filled in is not the same as right.
One idea about the findings rather than more findings — because it explains why almost none of this would appear on a fill-rate report.
Nearly everything this study found is filled in. The disk's hour field is filled in, at 57,220. Both of the tractor's hour fields are filled in, at 12 and 900. The former price is filled in, below the asking price. The snow blade's size is filled in, at 96 feet. A completeness check scores every one of those fields at a hundred percent, and every one of them gives a wrong answer. What catches them is not asking whether a field is filled but checking the value against the machine's own condition, its own model name, its own other fields.
Measured directly · EKOM's checks, not the analysis's findings
Every new machine is filed at a used-machine address.
Each listing's page, new or used, sits under the path the site uses for its used listings. The condition is in the record and the title; the path a shared link or a search result shows first says used. A platform decision with redirects attached, raised for the dealer and not counted among the 13.
Two more things were measured rather than found. More than three hundred machines carry no photograph, on a site whose own copy promises a picture of every machine. And the dealer's inside-sales phone number appears in two formats across the listings — with and without a hyphen after the area code — and is missing from 205 of them. A number a buyer taps on a phone should be one string.
What is strong is worth stating too. Stock numbers are unique across the entire inventory; the condition flag, the listing title and the listing address agree on new against used for every machine; every photograph carries its own machine's stock number in its file name. The record also already has the right fields: a secondary meter for a bale count, a reading field whose unit label elsewhere reads "Unit" and "Others", a model name on every blade named here that states the true width. Most corrections are derivations, not research.
Thirteen findings sounds like thirteen problems. Most of them read as five upstream mechanisms — and a mechanism, once confirmed, can be fixed once and stopped from recurring. Each tell below is a pattern in the data; the mechanism is EKOM's reading of it.
One meter field for every kind of meter
The tell: the hour field's own unit label is not always hours — an applicator re-read live shows a reading in "Unit", a snow blade in "Others". The vocabulary for "this is not hours" exists. On the balers the figure reads as a bale count under a label that says hours; on the disk it reads as belonging to no meter at all.
Two copies of the hour count, kept apart
The tell: all 31 disagreements are on machines flagged on order — consistent with two values captured at different moments and reconciled on arrival. Separately, 144 machines carry the displayed reading with no bare value behind it.
A size stored as a number, with the unit chosen by equipment type
The tell: all 33 wrong-unit sizes are blades, and the numbers cited are plausible widths in inches; the draper heads and sprayers in the same field are correctly in feet. It reads as a label chosen by type and a number typed by hand.
Industry assigned by equipment type, with types missing from the map
The tell: the 71 blanks fall into six equipment types rather than scattering across the inventory. A machine-by-machine omission would not be expected to cluster that way.
An identifier copied into the field beside it
The tell: all 14 serial-equals-stock machines are fabricated attachments from one maker, which carry no manufacturer serial. It reads as the stock number standing in for an empty field.
Correcting 33 blade widths by hand corrects 33 listings. Checking the unit against the equipment type at the point of entry corrects those and every blade listed after them — and the same check, applied to the meter field, is what would have stopped a bale count from becoming hours.
In McKinsey's 2024 survey of 3,942 business-to-business decision makers, 39 percent were willing to spend more than $500,000 in a single order through self-service digital commerce or remote online connections. The machines in this study are priced in that range — and the listing is what that buyer decides from.
A dealer group sells expertise as much as iron. The buyer on the phone is answered by someone who knows which of two tractors came back from rent, and whether a baler's meter counts hours or bales. Online, the fields carry that alone — across every store in the group, for the farmer three states away who reads the listing before calling, for the next dealership down the road quoting from it, and for the marketplace or assistant that copies it.
What this class of defect costs is not a missed click. A tractor with 12 hours in one field and 900 in another is a listing a buyer cannot trust either way. A clearance machine whose record says it was cheaper before reads as a markup wherever both prices show. A machine with no industry value is absent from the one filter built to find it. Each is a machine the dealer has or has coming, behind a listing that keeps the buyer from it.
A second clock runs underneath. The buyer comparing combines or dozers increasingly starts with an assistant rather than a search box, and an assistant reads the fields: hours, condition, price. On this inventory it finds a new disk listed at 57,220 hours and a tractor with two hour counts — wrong answers, and they do not correct themselves.
Three moves, in order. The first needs no new data at all — and the corrections that look most urgent deliberately wait for the dealer's answer.
This pass read and diagnosed. The same structural understanding powers what follows — turning a diagnosed inventory into one a buyer can choose a machine from with confidence.
1 · Apply the corrections the record already answers for itself
Convert the blade widths from inches to feet — the value divided by twelve, which the model name already states — leaving the draper heads and sprayers untouched. Fill the industry from the equipment type for the six types that are missing from the map. Clear the stock number from the serial field on the attachments that carry no manufacturer serial. Approved as patterns, not machine by machine.
2 · Settle the questions that gate the rest — starting with hours
Reconciling the two hour fields looks like a mechanical fix, and it waits: only the dealer's inventory system knows which reading a machine arrived with, and only the dealer can say whether a 999-hour demonstrator is new, whether six clearance prices were swapped at entry, and where a bale count belongs. EKOM lists every machine with both values beside it; the dealer answers once per pattern.
3 · Hold the line at intake
Keep ongoing intelligence where a machine enters the inventory, so the next bale count filed as hours, or the next width entered in inches, is caught the day the listing goes live rather than found later by a buyer sorting by hours.
This is how EKOM moves an inventory from insight to impact —
and keeps every machine described as itself as the inventory turns over.