Resolution Layer Case Study · Read & Diagnose
Industrial motors distribution.
A catalog read of one servo-motor category from an industrial distributor's web catalog — where about three in four item pages carry no specification table, and the rest sometimes disagree with themselves.
This is a real EKOM catalog analysis, with the client's identity removed. The client is a regional industrial distributor selling power transmission, motors and related supplies to maintenance, repair and OEM buyers. Its web catalog gives every orderable item number a page of its own: the item number, a short ERP line, a photo and, where it has been filled in, a specification table. EKOM chose one category, listed every item in it, read every item page, ran an automated read over them, and raised 11 candidate findings. 10 are reported here as actionable catalog defects. One was withdrawn as a modelling suggestion rather than a defect, and is in no count below.
What makes this catalog different is what the page is for. A servo motor is not chosen by its photo. A buyer chooses by torque, speed, frame, shaft and mounting, against a drive and a machine that already exist, and the item page is where those numbers are matched. When the table is empty the buyer is likely to call, or to go to a page that has one.
In most catalogs a missing field costs a click. Here it costs a call — or the order, to whoever's page has a table.
The manufacturer is not named here, and item numbers, photo file names and exact values have been removed or generalized, because on a distributor's catalog a manufacturer, model and defect together can identify the distributor. Percentages, ratios and the direction of every disagreement are exact; item counts are withheld.
What's inside
- At a glance — the severity split, where the findings concentrate, and how the category was read.
- What a buyer runs into first — most of the category with no specification table, and the line where the gap sits.
- The receipts — photos named for a different motor, and values that contradict their own page.
- One value, several names — what a facet filter does with a field that has no controlled list.
- Where it traces back, and what's next — the four mechanisms the findings collapse into, and the order to fix in.
Nothing was supplied and no internal system was touched. Everything below was read from the pages the distributor publishes on its own website, and the headline values were re-checked against the live pages.
10
Actionable findings
(11 flagged)
74%
Item pages with no
specification table
Where the 10 concentrate
Theme
What it is
Findings
No specification table
74% of the category has no table; the Series field is blank on 83%
2
Values that contradict the page
A unit that changes, a rating a digit off, a horsepower that disagrees with the kilowatts
3
A photo named for another motor
Dozens of items display a photo file named for a smaller frame
1
One value, several names
One mounting type in four spellings; one family in two names
2
Wrong category
A handful of DC drives filed under AC servo motors
1
Two identifiers, one field
A supplier part number and a barcode in a single field
1
How EKOM read this category
1
Read
The category as any visitor sees it: the listing pages, then every item page, one per orderable item number.
Every item listed was read, and every page loaded.
2
Profile
Infer what each field is for from how the pages themselves use it, then hold every item to that.
The item number and the ERP line are the keys each page is built on.
3
Analyze
Surface defects, rate severity, bind each to the items and values behind it — then recount against the archived pages.
10 findings, 6 themes, 4 causes.
Machine-surfaced signals from a cold read, not a verified defect list. The analysis's own counts ran low in several places, so every magnitude here is a recount from the archived pages. What remains was checked against the pages themselves, and the headline values against the live pages.
What a buyer runs into first.
The lead finding is about the part of the page a buyer relies on most: the specification table. In this category, for most items, there isn't one.
Critical · No specification table, no description
74% of the category's item pages carry no specification table and no product description.
A bare page carries a title, an abbreviated ERP line and a run of the manufacturer's order-code text. Power, speed, mounting and enclosure appear only as prose, shaft size not at all, and no field holds any of them, so there is nothing for a buyer to filter on or compare. A page with a table gives the same values as fields.
Field
A bare page
A page with a table
Horsepower
in the title; no field
a field
Speed
in the title; no field
a field
Shaft size
not on the page
a field
Mounting type
in order-code text; no field
a field
Enclosure
in order-code text; no field
a field
A person who reads the manufacturer's order codes can decode the bare page. A filter, a comparison table and an assistant cannot: they read fields, and for three of every four items in the category there is no table to read.
The gap is not spread across the category. It is one product line. 94% of the items in the largest line are bare, against 2% of the items in its sibling line — same manufacturer, same category, same site. The split follows the line, not the item, and that is what makes it fixable at the line.
Every item in the category, grouped by manufacturer line, with the share that has no specification table and no description. Item counts are withheld.
Line
Relative size
No table
Largest line
Largest
94%
Sibling line
About a quarter of the largest
2%
Special configurations
Small
92%
Other lines
Very small
45%
Whole category
All lines
74%
Critical · The same gap, seen in a filter field
The Series field is blank on 83% of the category.
Series tells a buyer which manufacturer family a motor belongs to. It is filled on about one item in six; on the rest, a Series filter cannot see the motor. The family is not hidden: on the two main lines it is the opening characters of the item number itself, so filling the field needs no outside source.
The same manufacturer's pages are nearly all filled on one line and nearly all bare on another. That reads like a catalog whose lines arrived by different routes — an inference from the pages, which the distributor's own records settle.
"No table, no description" means the page has no specification table and no product description block. Every bare item still carries its ERP line and the manufacturer's order-code text, which is why the page title can state a power or a torque even when no field does.
Photos named for a different motor.
Frame size is what a buyer checks a photo for. On dozens of items the photo the page displays is a file named for a smaller motor family.
High · The photo file is named for a different motor
Dozens of items display a product photo whose file is named for a different, smaller motor.
The same few photo files recur across many larger-frame items, which is what suggests a family-level attachment rather than a one-off. EKOM read the file names, not the pixels: a file named for a smaller frame on a larger-frame page is the signal, and whether the image itself shows the wrong frame is for whoever opens it to confirm.
Pattern
What the page displays
Verdict
Larger-frame item
a photo file named for a smaller family
disagree
Order-code variant of the same motor
its sibling variant's photo
normal, not counted
A shared photo is not the defect. Further items share a sibling variant's photo, which is ordinary for order-code variants of the same motor, and they are not counted. The rest are different: a buyer comparing flange, shaft and housing from the picture is comparing a different motor's. That the attachment is made by family is an inference from the file names, and the distributor's image library settles it.
Values that contradict their own page.
Three motors whose pages disagree with themselves. Each was checked on the live page, and the page says what the analysis said it says.
Critical · A unit that changes on the same page
A shaft listed in inches in the specification table, on a page whose own description says millimetres.
The table gives a shaft size with the unit "in"; the description on the same page gives the same figure in millimetres. A shaft that size in inches does not exist on a servo motor. The same page does it again: a temperature rating missing its lower bound, and a mounting code sitting in the Frame field while the Mounting Type field, where it belongs, has no value.
Critical · A rating a digit off
An ERP line that states a power rating ten times what the item's own description and table say.
Every sibling in the family reads one figure; this item's ERP line, which is also part of its page title, reads the same digits with one changed. The description and the table agree with the siblings. A search or a filter on kilowatts would place this motor in a power class roughly ten times too high.
Critical · Kilowatts and horsepower that disagree
A power rating in the ERP line and a horsepower in the table that differ by about a tenth.
Converted, the kilowatt figure gives a horsepower about 12% below the one the table prints. A sibling reads closer to the converted value. A second item carries the same table value but has no kilowatt figure to check against, so it is not counted.
One value, several names.
A facet filter only works if one thing has one name. Two fields give the same thing more than one, and two smaller findings sit alongside them.
High · One value, four spellings
One standard flange mount appears in the Mounting Type field in four different spellings.
All four mean the same standard mount. A Mounting Type filter splits one population four ways: even the most common spelling reaches only about half of the items that carry one of the four, and the rest of the category has almost no mounting value at all, which is the first finding again.
Medium · One family, two names
The same manufacturer lines appear in the Series field with and without a prefix.
On the larger line the prefixed form outnumbers the plain one by about four to one; on the other by about two to one. A Series filter on the plain form returns a minority of the family's filled items.
Two smaller findings
Four DC drives filed under AC servo motors, and four items with two identifiers in one field.
A few DC drives sit in the AC servo category with their motor fields blank. Separately, a few items hold a supplier part number and a barcode in one part-number field, so a lookup on either value may fail against the combined string.
Ten findings across a category of this size sounds like a long list of corrections. They collapse into four upstream mechanisms — and a mechanism can be fixed once and stopped from recurring. Each tell below is a pattern on the pages; the mechanism is EKOM's reading of it.
A product line loaded without its specification
The tell: 94% of one line is bare against 2% of its sibling: same manufacturer, same category, same site. The gap follows the line, not the item. A bare page carries an ERP line and order-code text and nothing else. It produced the bare pages and the blank Series field.
Values that look keyed or converted by hand
The tell: a unit that changes between description and table, a digit that changes between the ERP line and its siblings, a conversion that does not land. It produced the three contradicted pages.
Free-text values with no controlled list
The tell: one mount in four spellings, one family with and without a prefix. Nothing in the field stops a new spelling, so each load can add one.
Values attached at the wrong level, or merged
The tell: a photo file named for a smaller frame shared across larger ones; DC drives filed beside AC motors; a part number and a barcode keyed into one field.
Correcting the bare pages by hand corrects today's pages. Filling a line's specification from its source, holding each field to one controlled list, and checking each load before it goes live corrects those, and every page the next load would otherwise leave bare.
39%
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. A servo motor is a small line on that scale, but the principle holds at any size: the buyer who completes a purchase without a call is the buyer whose item page already answered the question.
McKinsey & Company, B2B Pulse survey, 2024, as reported by Digital Commerce 360, September 12, 2024
An industrial distributor's catalog sells technical certainty as much as inventory. The buyer on the phone used to be answered by someone at the counter who knew the motor. On a web page — and in whatever reads the page — the fields carry that alone. These findings are what happens when a page is assembled from a record that did not arrive intact. Nobody decided a motor should have no specification. It was never anybody's job to check the bare lines against the filled ones.
What this class of defect costs is not a sale — it is the call, or the order that went to a competitor's page. A buyer who meets a bare page is likely to call, or to go to a distributor whose page has a table. A buyer who meets a shaft in inches is likely to call too, and to trust the rest of the table less. A buyer filtering on mounting sees part of the motors that match. None of these is a decision the distributor made; each is a copy of what it knows that did not arrive intact.
A second clock runs underneath. The buyer pricing a replacement increasingly starts with a tool — a search filter, a portal, an assistant — and the tool reads the fields. It cannot decode a manufacturer's order code the way a person at the counter can. It takes the page at its word, and on these pages its word is sometimes missing and sometimes wrong. Those are wrong answers, and they do not correct themselves.
Three moves, in order. The first needs no new data at all — and the one that looks most mechanical deliberately waits.
This pass read and diagnosed. The same structural understanding powers what follows — turning pages a buyer has to second-guess into a record everyone reads the same way.
1 · Apply the corrections the pages already answer for themselves
Give every field one spelling and every family one name. Correct the contradicted values from the evidence on their own pages. Fill Series from the item number. Approved as patterns, not item by item.
2 · Settle the questions that gate the rest — starting with whether the record holds the specification
Filling the bare pages looks mechanical, and it waits on one answer per line: whether the distributor's record holds the specification the site never received, or whether the manufacturer's data is the source. So do the photos and the drives filed in the wrong category.
3 · Hold the line at the next load
Run the same read on the next batch of items before it is published — every table checked for presence, every value held to its unit and its siblings, every photo checked against its frame, every field held to one spelling — so these findings become a list of things that did not happen.
This is how EKOM moves a catalog from insight to impact —
and keeps every item page reading right as the business grows.