Resolution Layer Case Study · Read & Diagnose
Promotional drinkware.
A catalog read across the drinkware category of a branded-merchandise supplier that sells through distributors — tumblers, bottles, mugs and sets, where the page a distributor quotes from and the copy written for search engines and assistants are two different documents.
This is a real EKOM catalog analysis, with the client's identity removed. The client is a branded-merchandise supplier that sells through distributors and publishes a drinkware category on its own site. EKOM read every product page in that category, flagged 25 items, and reports 23 as actionable data defects. Two were set aside before the report: one described EKOM's own extract rather than the supplier's data, and one is a plausible value that only the supplier can define. Neither is in any count below. Every extent here was recounted by an analyst against the archived pages; the automated findings' own counts ran low and are not used.
What makes a drinkware catalog different is that each product page is read twice. A distributor reads the page — a name, a capacity, a set of dimensions, whether a lid is in the box — and quotes from it. A search engine, a shopping listing or an assistant reads a second copy, the structured data written beneath the page for machines, and takes it at its word. When the two copies agree, the page is dependable. When they disagree, or a field disagrees with the product's own name, the record simply says two things.
A distributor's relationship with a supplier runs through the product page. The facts a rep quotes from it — and the characters a machine reads it in — are only as good as the field that holds them.
Product names, item numbers, brand names, partner names and the supplier's own category vocabulary have been removed or generalized here, because in a promotional catalog a brand roster and a product line together identify the supplier. 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.
- The copy machines read — a description with its punctuation removed, a brand field that is blank on the supplier's own products, and an offer a listing may not parse.
- A 32 oz mug, specified as 20 — capacity that contradicts the product's name, and capacity that is missing.
- A tumbler 433 inches wide — three impossible dimensions, and five incomplete ones.
- Where it traces back, and what's next — the four mechanisms the findings read as, and the order to fix in.
Nothing was supplied and no internal system was touched. Everything below was read from the supplier's public product pages, and every product named in the findings was re-read live the same day.
23
Actionable findings
(25 flagged)
Where the 23 concentrate
Theme
What it is
Findings
Machine description stripped
The structured-data description has its punctuation removed on all but one product
1
Machine brand and offer incomplete
Blank brand on 233 house-brand products; an offer a parser may not read
2
Capacity contradicts the name, or is missing
A 32 oz mug at 20 oz; 52 products with no capacity attribute
2
Dimensions impossible or incomplete
A tumbler 433 inches wide; two cases of 148 and 118 inches; five incomplete records
5
Lid fields contradict each other
Included and optional at once; included with no value; "No" beside a closure
3
Insulation missing or typed inconsistently
Hours in the copy and not in the fields; one insulation typed three ways
5
Non-drinkware items and values that do not apply
Empty gift boxes and a spare part listed; a powder coating on a cardboard box
2
Placeholders and unlikely values
"N/A" as a value on 15 products; a single-wall bottle typed "Air"
3
How EKOM read this catalog
1
Read
Every product page in the category, fetched the way a search engine fetches it — no file handoff, no credentials, no integration.
One row is one sellable item, keyed by item number; the category's listing reconciles to the pages read.
→
2
Profile
Hold what each product says about itself against itself: its name, its copy, its specification block, its case data and the structured data written for machines.
A product's own name and copy are the answer key for most of what follows.
→
3
Analyze
Surface defects, rate severity, and bind each one to the products and values behind it.
23 findings, 8 themes, 4 causes.
Machine-surfaced signals from a cold read, recounted by an analyst — not a supplier-verified defect list. Some will prove intentional, which is why each is tied to the values behind it.
Every product page carries a record for search engines and AI shopping agents. It is not the page a distributor reads, and on three fields it is wrong in the same way on nearly every product.
Critical · The machine description has its punctuation removed
On all but one product, the structured-data description is the page's text with the punctuation stripped — and "18/8 stainless" reads as "188".
The description in each product's structured data was compared with the description on its page. On every product but one the words are the same and the characters are not: hyphens, slashes, percent signs, apostrophes and symbols are gone and nothing replaces them. On the remaining product the structured-data description is empty. Hyphens drop out on 347 products, percent signs on 91, slashes on 61, and apostrophes on about 150.
What the page says
What the structured data says
Verdict
Double-wall 18/8 grade stainless steel
Doublewall 188 grade stainless steel
meaning changed
won't · you're · leak-proof · non-slip
wont · youre · leakproof · nonslip
words changed
100% recycled
100 recycled
unit lost
High · The machine brand is blank on the supplier's own products
233 house-brand products declare no brand — and every partner-brand product declares one.
The brand field in the structured data is filled on every partner-brand product and empty on every product carrying one of the supplier's two house brands, which split almost evenly between the two. A listing or an assistant asked "which brand is this" gets a blank for what the supplier makes and a name for what it distributes.
High · The machine offer is not in a form a listing can read
Availability is declared on every offer re-read; the price sits under names and values a standard parser does not expect.
On the 40 pages re-read live, every offer is a price range carried under lower-case keys where the vocabulary uses mixed case, with no single-price key, and a currency field that holds a language-and-region code rather than a currency code. One of the 40 has an empty price. The price is on the page that visitors see; whether the structured data is meant to carry it is a question for the supplier.
The uniformity is the tell. A defect that touches nearly every product in the same way does not read as a set of typing errors; EKOM reads it as one template doing one thing, and a correction made once would hold.
A 32 oz mug, specified as 20.
Capacity is the first thing a distributor's customer filters on. The page carries it twice — in the product's name and in a specification field — and on these products the two disagree.
Critical · The capacity field contradicts the product name
Four products carry a capacity that disagrees with their own name — one by 12 ounces.
An insulated mug named as 32 oz carries a capacity field of 20 oz. A slim bottle named as 15 oz carries 20 oz; a straw tumbler named as 15 oz carries 14 oz; a travel tumbler named as 15 oz carries 16 oz. Two more differ only by a rounding — 13.5 to 13, and 16.9 to 17. Which of name and field is the truth is a question only the supplier's spec sheet answers.
High · No capacity field although the name states one
52 products state a capacity in their name and carry none in the specification — so a capacity filter cannot find them.
The category pages offer a capacity filter. A product without the field is not in it. The automated read counted 25 such products; the analyst's recount found 52. Three products re-read live have a specification block that is empty altogether, so capacity is missing along with every other spec. The corrected value is the one already in the product's name, where name and field do not disagree.
A distributor whose customer asks for a 32 ounce mug is given a specification that says 20, and resolves it by asking someone. Every such question is time taken from the relationship the page is supposed to support.
A tumbler 433 inches wide.
Product and shipping-case dimensions feed freight quotes and packaging decisions. Three values in the category are physically impossible, and they are the only three.
Critical · Physically impossible dimensions
A 20 oz tumbler listed as 433 inches wide; two shipping cases listed as 148 and 118 inches.
A recycled tall tumbler's product width reads 433.00 inches, while its own shipping case is under two feet in every direction and a sibling tumbler reads about 3.35 inches wide. A 25-unit case weighing 18 lb has a width of 148.00 inches; a 36-unit case weighing 11 lb has a depth of 118.00 inches, where comparable cases sit at 18 to 19. Every other dimension in the category is plausible, which points to typing errors and not to a unit problem. A freight quote built on a 118-inch case is wrong by the same factor.
High · Dimensions incomplete
Four products show two of their three dimensions, and one shows none.
Three products give height and width with no depth. A fourth is the mirror image — height and depth, no width. And a boxed tumbler has no product dimensions and an empty shipping-case block, although it carries ten other specifications. For a round bottle a missing depth could be derived from the width; whether a derived figure may be published as if it were measured is the supplier's call.
The sibling products in the same category give the likely scale of each impossible value, and the supplier's spec sheet gives the real one. EKOM holds the three for confirmation instead of correcting them.
Lids and insulation: the fields disagree.
Whether a lid comes with the product, and how long it holds a temperature, are the two facts a buyer checks. On these products the page's own fields contradict each other.
High · Lid fields contradict each other
The lid is "included" and "optional" on 10 products, "included" with no value on 20, and "No" on 7 products that have one.
Ten products list both an included lid and an optional lid in their features. Twenty list an included lid in their features and carry no value in the field built to say so; the automated read counted 6 and 3, and the analyst's recount found 10 and 20. On seven gift sets and bottles the lid field reads No beside a feature that says a lid is included and a closure field that says it twists on. Whether "No" describes the bottle or the gift box is a question the page does not answer.
High · Insulation facts missing, or typed three ways
Hours stated in the copy are absent from the fields, and one insulation is written three different ways.
A pair of tumblers states its cold hours in its key features and carries no insulation field. Another carries cold and hot hours in its description and neither in the specification; two more state hot hours in their copy and have no hot field. Separately, one kind of insulation is typed with its full name on one product, shortened to Vacuum on two, and not at all on three. A filter on one spelling finds the products typed that way and misses the rest.
None of this is hidden. Every fact is on the page — in the name, in the features, in the copy — and the field built to carry it is empty or says something else. Most corrections are derivations, not research.
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.
Much of what this study found is filled in. The description is filled in, with its punctuation removed. The width of the tumbler is filled in, at 433. The opening type is filled in, with "N/A". The coating of a cardboard gift box is filled in, with "Powder". 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 product's own name, its own copy and its own other fields.
Medium · Defaults standing in for values
"N/A" is a value on 15 products. A single-wall plastic bottle is typed "Air". A gift box has a powder coating.
The opening-type field reads N/A on 15 products and is blank on another 111, so a facet filter offers "N/A" as if it were a kind of opening — and one of the 15 is a bottle with a screw-on lid. Three empty gift boxes and a replacement part are listed as products, and with two coaster sets they carry a coating of Powder, a drinkware attribute that does not apply to them. Each reads like a default carried onto a record that was never meant to have the value.
What is strong is worth stating too. The category listing and the pages read reconcile exactly, and every item number is unique. Every product has a materials statement and a description. The partner-brand products fill the machine brand field correctly, which proves the field works. And the category contains its own answer key: a sibling tumbler with a sane width, a sibling case with a sane depth, a sibling bottle with the insulation typed the right way.
Twenty-three findings sounds like twenty-three problems. Most of them read as four 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.
A separate template writes the machine copy
The tell: the stripped description, the blank house-brand field and the offer are each uniform across the category — the same behavior on every product, which reads as a template and not as people typing. Three findings, one rule.
Two sources disagree, and nothing referees
The tell: the same fact is entered in two places — the name and the capacity field, the features and the lid field, the copy and the hours fields — and the two differ. Ten findings read this way, and the page shows nothing that compares the copies.
A default fills in where a value should be
The tell: "N/A" in a facet, a powder coating on a cardboard box, "Air" on a single-wall bottle, and gift boxes and a spare part carrying the category's attributes. Each value reads like a default carried in to avoid a blank. Five findings.
Typed values with no plausibility or completeness check
The tell: a width of 433 inches, case sizes of 148 and 118, four dimension strings with one of three missing and a record with none. The values shown suggest no check at entry for a number that cannot be real. Five findings.
Correcting one tumbler's width corrects one tumbler. A check that refuses a 433-inch width corrects it and every width entered after it — and the same template rule that restores a description's punctuation restores it on every product at once.
In Gartner's survey of 632 business-to-business buyers, 69 percent reported inconsistencies between the information on a sales organization's website and what its sellers told them, and 61 percent preferred an overall rep-free buying experience.
A supplier that sells through distributors sells reliability as much as product. A rep answers the phone and knows which tumbler runs large and which case ships heavy. Online, the fields carry that alone — for the distributor quoting from the page, for the customer behind the distributor, and for the search engine or assistant that copies it.
What this class of defect costs is not a missed click. A mug whose specification says 20 ounces under a name that says 32 is a listing a buyer cannot trust either way. A capacity filter that cannot find 52 products is a sale that never reaches them. A description that reads "188 stainless" is a spec no search engine can match. A freight quote built on a 118-inch case is wrong by the same factor.
A second clock runs underneath. The buyer comparing drinkware increasingly starts with an assistant rather than a search box, and an assistant reads the record, not the page: the description, the brand, the offer. On this catalog it finds a spec with its characters removed, a blank brand on the supplier's own products and a price in a form it may not recognize — 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 supplier's answer.
This pass read and diagnosed. The same structural understanding powers what follows — turning a diagnosed catalog into one a distributor, a search engine and an assistant can all read the same way.
1 · Apply the corrections the record already answers for itself
Write the machine description from the page description with its punctuation intact. Fill the capacity from the name on the products that have none. Move the hours stated in the copy into the hours fields, and type each insulation one way. Approved as patterns, not product by product.
2 · Settle the questions that gate the rest — starting with capacity
Where a name says 32 and a field says 20, only the supplier's spec sheet knows which is true. The same goes for the three impossible dimensions, for what "no lid" means on a gift set, for what a cold-retention time means on a single-wall product, and for which price the machine record should carry. EKOM lists every product with both values beside it; the supplier answers once per pattern.
3 · Hold the line at entry
Keep ongoing intelligence where a product enters the catalog, so the next width of 433 inches, or the next "N/A" in a facet, is held for review the day it is entered rather than found later by a distributor.
This is how EKOM moves a catalog from insight to impact —
and keeps every product described as itself as the catalog grows.