Resolution Layer Case Study · Read & Diagnose
Electrical tools & test equipment.
A catalog read across a sample of the tools and test-equipment product pages of an electrical distributor — where a buyer chooses an instrument on its type, its voltage and its safety category before anything else.
This is a real EKOM catalog analysis, with the client's identity removed. The client is an electrical distributor selling tools, test instruments and tool storage through a public online catalog. EKOM read a sample of 117 of its product pages, raised 13 findings, and reports 11 as actionable catalog defects. Two were withdrawn after re-checking and are in no count below. The extents of 20, 23, 14 and 7 below were recounted by an analyst against the saved pages, and where the automated count differed — the brand-only and color findings both ran low — the recount is used.
What makes a tools and test-equipment catalog different is that the specification is the product. An electrician choosing a cable certifier filters on its type, an insulation tester on its safety category, a clamp meter on its measurement range. When those fields are missing, the product does not look incomplete; it is simply not in the list the buyer is looking at. And when a field holds the wrong kind of fact, the same filter quietly compares things that are not alike.
A product page that is wrong misdescribes a product. A page that says almost nothing does something quieter: it takes the product out of every filter the buyer starts in.
Item numbers, manufacturer and product names, the distributor's category vocabulary, its page text and the exact sitemap figures have been removed or generalized here, because on a distributor's catalog a maker, a product and a category together identify the distributor. Counts are exact within the sample, and where quoting a field value would identify a page, the value is described instead of quoted.
What's inside
- At a glance — the severity split, where the findings concentrate, and how the read was done.
- What every page tells a search engine — four checks that held on every page measured, and a sitemap that offers only part of the catalog.
- Instruments with a brand and nothing else — 20 pages whose specification table has one row, and one page that carries three brands.
- Misspelled values, wrong shapes, and one fact entered two ways — what a buyer typing the right word or filtering on the right value does not reach, plus a voltage field with four meanings and two unfinished records.
- 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 distributor's public product pages and its sitemap — the sitemap six days before the pages — and prices, which sit behind a login, were not seen.
11
Actionable findings
(13 flagged)
Where the 11 concentrate
Theme
What it is
Findings
Spec tables with only a brand
20 pages whose specification table has a single row, Brand
1
One product, three brands
A title, a heading and a specification table naming three different makers
1
Misspelled values
Type, count and name fields with a letter wrong
3
Values in the wrong shape
A temperature range in a size field; a dimension with a unit on the last value only
2
One fact entered two ways
Multi-color values written two ways; drive sizes with no unit
2
One field, four meanings
A voltage field holding four different kinds of voltage
1
Unfinished records
Two products with no brand and a warehouse line for a name
1
How EKOM read this catalog
1
Read
117 live pages, drawn across 25 subcategories weighted to each subcategory's size and read in an ordinary browser at human pace — plus the product sitemap. No file handoff, no credentials, no integration.
A sample of pages supports patterns and examples, not a census.
→
2
Profile
Infer what each field is for from how the catalog itself uses it, then hold every page to that — a brand against its title, a size against its unit, a color against its siblings.
Most findings test a value against its neighbors.
→
3
Analyze
Surface defects, rate severity, and bind each one to the pages and values behind it.
11 findings, 7 themes, 5 causes.
Machine-surfaced signals from a cold read of a sample, re-checked by an analyst — not a distributor-verified defect list. Some will prove intentional, which is why each is tied to the values behind it. The catalog changes daily; a page described here may have been corrected since.
What every page tells a search engine.
Four checks EKOM ran on every page the sample measured. They are measurements, not analysis findings, and they are not counted among the eleven. The same result on every page is what a page template does, and what a person typing does not.
| Check | What the pages show |
|---|
Structured product data
the markup search engines read | No JSON-LD product markup on any page measured. The check looked for JSON-LD only, so no page hands a search engine its brand, its specifications or its identity in that form. |
| Unrendered template text | A placeholder in curly braces shows on every page measured, where a heading should print its name. |
Meta description
the line a search result shows | On every page measured it is the manufacturer number followed by the product name, with no sentence about what the product does or who it is for. |
| Dead address in the sitemap | One sitemap address opened was a 404 — an address that no longer resolves. |
Most pages already show a title, a brand and a specification table. What is missing is the layer that hands them to a search engine — and a template that prints a placeholder and a template that omits structured data are each one fix that reaches every page.
Measured · Sitemap
The sitemap is a truncated view of the catalog.
The sitemap stops short of the catalog rather than reaching its end. The skew shows it: some whole departments get a single product address each, one department gets thousands, and a large block of addresses sits under a discontinued-products branch. Search engines are offered only part of the catalog, weighted toward discontinued items and one department, and a product that is not in the sitemap is found later, or not at all. The true catalog size was not measured. A sitemap built from the live catalog, rather than from a generator that stops early, is the fix.
Instruments with a brand and nothing else.
The lead finding is about the products where a buyer most needs the specifications, and where a filter cannot find them.
Critical · Specifications missing
20 pages in the sample carry a specification table whose only row is Brand — among them a cable certifier.
Eighteen of the 20 are instruments from one test-equipment maker and two are carts from one cable maker. The instruments include a cable certifier whose page title is only the maker's name and a category, optical loss test sets, an insulation tester, clamp meters, an EV-charging station tester and pressure modules. On the pages the analysis cited, the specification table holds only the brand.
An electrician choosing a cable certifier or an insulation tester filters on type, voltage and safety category. These pages carry none of the three, so they drop out of the type, voltage and safety-category filters and the comparisons built on them, while products with more specification fields stay in. On a test instrument the missing fields are the whole decision.
What the analysis proposes to fill — from the manufacturer's own documentation
Field
What it should hold
Pages cited
Type
The product type, such as Clamp Meter, Insulation Tester, Cable Tester or Cable Cart
11
Voltage
The operating or measurement voltage
8
Safety category
The safety category rating, such as CAT III or CAT IV (measurement category)
6
Display
The display, such as Backlit or Color LCD
7
Amperage
The current measurement range or rating
2
Counts. The table counts the 11 pages the automated analysis cited. The analyst's recount against the saved pages finds the same state on 20. The distributor confirms which fields it wants shown for each instrument class; the values come from the manufacturer.
One product, three brands.
Brand is the field a buyer filters on first. On one page it takes three values at once.
Critical · Conflicting brand · Read from the rendered page
A case whose title names one maker, whose heading names another, and whose specification table names a third.
One compartment case carries three brands: the browser-tab title begins with one maker, the heading and the category name a second, and the specification table's Brand names a third. Its Includes field, where the accessories belong, holds a raw image address from the case maker's own site.
Field
What the record holds
Verdict
Product name
Names the second maker
agree
Category
Names the second maker
agree
Browser-tab title
Begins with a first maker's name
conflicts
Brand
Names a third maker
contradicts name
Includes
A raw image address, where accessory text belongs
wrong content
A purchasing agent who sees one maker in the title and another in the specification table stops trusting the rest of the page. The analysis reads the name and the category as the right answer and proposes clearing the image address from Includes; the distributor confirms the brand. This page was read from the rendered page and not from a saved copy.
Misspelled values, and values in the wrong shape.
Five findings, one effect: a buyer typing the right word, or filtering on the right value, does not reach the product. Every row is a real field on a real page, and for most of them the correct form already appears elsewhere in the catalog.
Field
What it actually holds
What it should hold
Misspelled values — three findings · High
Type of a drill bit set
A letter wrong in the type
Spelled correctly
Type of a hand-tool set
A word misspelled inside a long list of tool types
Spelled correctly, the rest as written
Type of a multimeter
A letter wrong in the voltage type
The form the sibling multimeters use
Count field and name of a multimeter
A letter wrong in the same word, in both fields
The form a sibling multimeter uses
Name of a screwdriver set
A letter wrong in the name
The spelling a sibling set uses
Values in the wrong shape — two findings · High
Size of a temperature probe
A temperature range, in the field built for a physical dimension
A dimension, or nothing — the range is a different attribute and the page has no field for it today
Size of a multimeter
Three dimensions with the unit on the last one only
The unit on every dimension
These are derivations, not research. A misspelled word in a name or a type breaks keyword search and creates a facet label of its own: a buyer who types the right spelling does not find the product whose name has a letter wrong, and a misspelled type sits in a filter beside the correct one. For most rows the correct form is already in the catalog or on the page — which is why a pattern, checked once, corrects every record that matches it.
One fact entered two ways, and records left unfinished.
Four more findings. None of them is a wrong answer; each one splits a single fact across two filter values, merges different facts into one field, or leaves a record in the state an enterprise resource planning (ERP) system holds it.
Medium · Inconsistent format
In this sample, multi-color values are written two ways: run together on 23 pages and slash-separated on 7.
Multi-color values are run together with no separator on 23 pages — one two-color pair on 14 of them and other two- and three-color pairs on the rest — and slash-separated on 7. The same combination splits across two filter values. Separately, all 4 drive-size values in the sample carry no unit: half-inch on two pages and quarter-inch on two. A buyer who filters for black-and-red tools meets two lists, and a buyer who filters for a half-inch drive meets a value with no unit to match.
Medium · One field, four meanings
Voltage means four different things.
The Voltage field holds the maximum measurable voltage on multimeters (1000 V), a range on one, a rating in kilovolts on a live-line tool, and the battery voltage on a cordless impact wrench (18 V). Inside one subcategory this is usually harmless; in any comparison across categories it is misleading, because a buyer filtering for 18 V tools and a buyer filtering for 1000 V meters are reading the same field for different facts. The analysis recommends two fields in place of one — an operating voltage and a measurement voltage range — which is a field-design decision for the distributor.
Unfinished records · No brand, an ERP line as the name
Two products read like warehouse entries.
One product is named with a stock code followed by an abbreviated description in capitals; the other is named the same way. Neither carries a brand. Both sit in tool-storage subcategories beside items from one maker, which suggests the brand, but that has to come from the source data and not from a neighbor. A customer-facing name follows the catalog's own convention — brand, product line, type, key attributes — and a record with no brand is invisible to the brand filter. The analysis marked this finding at lower confidence for that reason, and it is reported without a severity rating.
Eleven findings and four site-wide checks sound like fifteen problems. It is fewer. Read together, most of them read as five upstream mechanisms — and a mechanism, once confirmed, can be fixed once and stopped from coming back. These are readings from a sample, not diagnoses.
One page template, three symptoms
The tell: the unrendered placeholder, the missing JSON-LD markup and the part-number-and-name meta description are identical on every page measured. A person typing does not produce the same result on every page; a template does.
A manufacturer data load that reached some records and not others
The tell: eighteen of the 20 brand-only pages are from one instrument maker, yet other pages from that maker in the sample carry specification fields beyond the brand. The gap follows the record, not the manufacturer.
A sitemap generator that stops early
The tell: some departments get one product address each while one department gets thousands.
Free-text entry with no check against the catalog's own vocabulary
The tell: most misspellings have a correct sibling elsewhere in the catalog. Color is entered two ways, the unit is missing on every drive size, and two records carry a warehouse line for a name.
One field made to carry several meanings
The tell: Voltage holds four different facts, Size holds a temperature range on one probe, and Includes holds an image address on one case. In each, a field built for one thing is holding another.
Readings, and two open questions. Each mechanism is a reading of a sample. Why one product carries three brands, and whether the 20 brand-only pages are a data-load gap or deliberate, cannot be read from the pages — the data-load reading stays a hypothesis until the distributor answers, and the report puts both questions to it rather than printing a cause.
Correcting 23 color values by hand corrects 23 pages in this sample. Checking a color against one convention at the point of entry corrects those and every product added after them — and a template and a sitemap generator are each changed once and reach every page.
Worth stating before the next page, because every fix depends on it.
The right answer is already on the page next door. Most misspellings in this study have a correct form on a sibling item. A correction that is a derivation from the catalog's own records needs no outside research and carries little risk of a wrong answer.
The catalog already knows how to describe a product. The sample spans 25 subcategories, and the vocabulary for a clamp meter is not the vocabulary for a drill bit; the catalog already makes that distinction field by field, which is a good foundation for a consistent product record.
The product pages are published for search engines. The distributor's robots.txt allows product pages, and the sitemap lists some of them. The product data does not need to be unlocked; it needs to be handed over in a form search engines and assistants read.
The catalog has the right fields and, beside most mistakes, the right answer. What is missing is the layer that checks one against the other — and that is a problem that can be fixed once.
In McKinsey's 2024 B2B Pulse survey of nearly 4,000 business-to-business decision makers, one in three said that at any given stage of the buying journey they want digital self-serve options, and buyers used an average of ten interaction channels along the way. A product page is the channel that is always open.
For a test instrument, the fields are the answer — which safety category it carries, whether a voltage is a working rating or a measurement range. They are what the electrician reads before ordering, what the search engine indexes, and what an assistant copies.
The cost follows the field. A cable certifier with a brand and nothing else is absent from the type, voltage and safety-category filters a buyer starts in and from the comparisons built on them. A misspelled type or name is missing from the search a buyer types. A color entered two ways splits one filter into two. A product with three brands tells a purchasing agent the page cannot be trusted on the others. Each is a product on a page that keeps the buyer from it.
A second clock runs underneath. Buyers increasingly start with a search engine's summary or an assistant, and both read structured fields. On the pages measured they find a placeholder and no JSON-LD product markup, so the pages are unlikely to qualify for the rich results that read it.
Three moves, in order. The first needs no new data at all — and the corrections that look most urgent deliberately wait for the distributor's answer.
This pass read and diagnosed. The same structural understanding powers what follows — turning a diagnosed catalog into one a buyer can choose a product from with confidence.
1 · Apply the corrections the record already answers for itself
Give each misspelled word its correct spelling, taken from the sibling item where one exists. Split every run-together color at the capital letter and join it with a slash, the form already on seven pages. Put a unit on every drive size and on every dimension. Approved as patterns, not record by record.
2 · Settle the questions that gate the rest
Five questions wait for the distributor: where the product data is generated and why the pages print a placeholder; whether the short sitemap is deliberate; where the specifications for the 20 brand-only pages in the sample live; which brand is right on the three-brand record and the two with none; and whether Voltage should be split in two. EKOM lists the brand-only pages and prepares the specification fields that apply to each from the manufacturer's published documentation; the distributor confirms which fields to show per instrument class.
3 · Hold the line at intake
Keep ongoing intelligence where a record enters the catalog, so the next misspelled type or run-together color is caught the day the product is listed rather than found later by a buyer.
This is how EKOM moves a catalog from insight to impact —
and keeps every product described as itself as the catalog grows.