Resolution Layer Case Study · Read & Diagnose
Mattresses & bedding.
A catalog read across the mattress assortment of a large home-furnishings retailer — in the one furniture category bought by a single word, where that word is carried by two numbers that have to be the right way round.
This is a real EKOM catalog analysis, with the client's identity removed. The client is a large home-furnishings retailer that sells mattresses under its own brands alongside the national manufacturers it resells. EKOM read its live public mattress catalog and its sitemap, and surfaced 17 flagged items, 16 of which are actionable catalog defects. The seventeenth — a handful of prices out of line with their own size ladder — was raised at low confidence, was not carried through the analyst's re-derivation, and is in no count here.
What makes this category different is how the shopper decides. Nobody buys a sofa by one word. A mattress is bought by its size first — Queen, King, California King — and everything else is chosen inside that. In a showroom the size explains itself. Online it is two numbers, width and depth, and they do more work than any other field on the page: the size filter, the delivery quote, the comparison against another retailer, and increasingly the assistant that recommends a mattress before anyone visits a store.
A wrong attribute usually costs a click. A mattress bought on the wrong dimensions costs a delivery, an unwrapping and a return — the most expensive way a mattress sale can end.
The findings are reported as facts of the record — what each field holds, set against what its own siblings hold. Product-field claims were computed from the harvested product table; the page-level data behind that table was not archived on this run, so they are strong inference rather than proof. Sitemap claims were recomputed from sitemap files archived on the day of the read, and are firm.
On a multi-brand retail catalog the brand roster is itself a fingerprint, so every brand, product-family and series name has been removed, along with item numbers, prices, dates, category labels and URL patterns. Standard mattress footprints are given in round inches; the counts are real.
What's inside
- At a glance — the severity split, where the findings concentrate, and how the read was done.
- What a shopper runs into first — a width-and-depth reversal that follows brand ownership almost perfectly.
- The receipts — a series in which every listing carries another bed's measurements, groups with dimensions missing, and products that cannot be bought.
- What the machines read — one product at two addresses, names that identify nothing, and a sitemap that cannot tell a dead page from a live one.
- Where it traces back, and what's next — the six upstream mechanisms these findings collapse into, each with the tell that proves it, and the order to fix in.
Nothing was supplied and no internal system was touched. Everything below was read from the public catalog.
16
Actionable findings
(17 flagged)
Where the 16 concentrate
Theme
What it is
Findings
Dimensions that do not match the size
Width and depth reversed; a series carrying other sizes' footprints; groups with dimensions missing; one size thicker than its family
5
Category and catalog scope
An internal label on a live listing, one-off category values, and non-mattresses inside the mattress catalog
3
Names that do not identify the product
Many products under one name, and a series with the size left out
2
Sitemap signals
Dead pages served as live ones; one priority value on every product
2
One product, two addresses
The same product published at two URLs
1
Listings that cannot be bought
No price and no availability
1
Missing brand
Listings absent from brand navigation
1
Specifications that cannot be queried
No field for size, firmness or construction
1
Six findings are rated critical and ten high. Three of the high findings were measured by EKOM's analyst from the archived sitemap and product table rather than raised by the automated pass.
How EKOM read this catalog
1
Read
The public catalog and its sitemap as any shopper, marketplace or assistant sees them — no file handoff, no credentials, no integration.
A large slice of live mattress listings, plus the full sitemap.
→
2
Profile
Infer what each field is for from how the catalog itself uses it, then hold every product to that.
The national brands set the convention: width side to side, depth head to foot.
→
3
Analyze
Surface defects, rate severity, and bind each one to the products and values behind it.
16 findings, 8 themes, 6 causes.
These are machine-surfaced signals from a cold read, checked by an analyst — not a verified defect list. Some of what a read like this finds turns out to be intentional. That is why every finding is shown with the value behind it wherever one exists, and why the ones needing a judgment only the business can make are kept separate.
What a shopper runs into first.
The largest finding in the analysis is also the most orderly. It is not scattered across the catalog. It follows ownership.
Critical · Dimensions that do not match the size
374 mattress listings publish the head-to-foot length as the width.
The catalog's own convention, followed by the national brands, is width side to side and depth head to foot. On these listings the two are reversed. A queen is sixty inches wide and eighty long; these records say the opposite.
Listing
What the record holds
Reads as
An own-brand queen
width 80 · depth 60
sideways
A national-brand queen
width 60 · depth 80
correct
Brand group
Listings reversed
Pattern
The retailer's own brands — every one of them
99 to 100% of each brand's listings
near-total
National brands the retailer resells
none at all on four; at most about one listing in eight on any one brand
exception
Hundreds of people typing dimensions do not agree to be wrong along a line of corporate ownership. A mapping does — one transformation between the owned brands and the storefront that writes length into width.
That makes the largest finding in the report the cheapest one to fix. It is most likely correctable in one place per affected feed rather than listing by listing — and the convention to correct toward already exists inside the same catalog, on the brands it resells. The national-brand feeds prove the storefront, the fields and the read can all carry dimensions the right way round. The swap is not a platform limitation.
Measurements that belong to another bed.
A reversed record still describes the right mattress, sideways. These do not. Each publishes a dimension pair that belongs to a different size entirely.
Critical · Dimensions that do not match the size
In one national-brand series, every listing carries another size's dimensions. None is correct.
Across both sub-series, the size in the name and the size in the numbers never agree. A customer ordering a California King is shown Queen measurements.
Stated size
Published footprint
Belongs to
California King
about 60 × 80
Queen
California King
about 38 × 74
Twin
Full
about 76 × 80
King
King
72 × 84
California King
Queen
76 × 80
King
Twin XL
about 53 × 74
Full
Six of the ten listings shown. Every published value is a valid mattress size — just not the one in the name. Nothing is garbled; it is misplaced, which means the correct pair for each listing is known.
Critical · Category and catalog scope
One live listing is half a bed, measured as the whole one.
It is named as a California Twin — one half of a split California King, roughly 36 to 38 inches wide. Its dimensions, 72 × 84, are the whole California King, twice the width of the half the name describes. And its category is an internal label meant to stay hidden. Read together, the two suggest the component record of a split set, published by mistake.
The same failure appears on split sizes. Seven listings named Split King publish 84 × 72 — a California King's footprint, with the axes reversed as well: two errors in one value. Three named Split California King publish a depth of 79.5, the standard length, where a California King is 84 long.
Missing where it matters most.
Three findings where the record is not wrong so much as absent — each a tightly bounded group whose siblings are complete, which is what makes each one look like a single broken mapping rather than a scatter of omissions.
Critical · Listings that cannot be bought
Five mattress listings publish no price and no availability.
Four of them sit in one own-brand series whose other listings all carry a price and a stock status; the fifth is a national-brand listing. They cannot be added to a cart, and any shopping feed that validates the record rejects them — Google's product listings require both fields. Their names carry no size at all: the inch figures in them are thickness.
High · Dimensions that do not match the size
Fifteen premium split California Kings have a width and nothing else.
All fifteen publish a width of 36 and leave height and depth empty. One brand, one size, weights present on every one. Anything that reads thickness — a filter, a feed, a comparison engine — finds nothing to read. Their non-split counterparts carry full heights, and the height of each split listing is already written into its own product name.
Critical · Dimensions that do not match the size
One size in a six-size family is five inches thicker than the rest.
The California King publishes a height of 16. Every other size of the same mattress — Twin through King — publishes 11. In any thickness filter, one size of one mattress appears to be a different product. It was findable at all only because the other five agree.
A catalog has two ways of saying which product is which — the address it publishes and the name it shows — and a third layer, the sitemap, that tells crawlers what exists. All three break here.
Critical · One product, two addresses
Forty-one mattresses are published twice.
Each appears once at a size-specific URL and once at a second, master-style URL — both listed in the sitemap, identical on every field compared. To a search engine that is two competing pages splitting whatever ranking signal the product earns. To anything that counts or syncs the catalog, it is two products where there is one.
High · Names that do not identify the product
Forty-eight product names are each shared by more than one product.
119 listings carry a name that is not unique. The largest cluster is twelve distinct products under one identical name, with nothing in it to tell them apart — their non-split siblings carry series and firmness in the name, and the split versions lost it. In another series, sixteen listings leave the size out of the name entirely while the Queen alone is named correctly.
Layer
What it holds
Effect
Structured fields
no field for size, firmness or construction — all three live only in the name
unfilterable
Sitemap status
three listed mattress URLs return a "No Results" page with a success code, not a 404
dead as live
Sitemap dates
those dead URLs carry the same last-modified date as hundreds of live products
indistinguishable
Sitemap priority
one value, 0.5, on every product URL in the whole sitemap
no signal
The structural finding is the one that ties the report together. To filter on size, anything reading this record has to infer it from the dimensions — reversed on every brand the retailer owns — or parse it from a name that, on sixteen listings, does not contain it. Any size filter built on this record rests on its two least reliable surfaces.
16 findings sounds like 16 problems. They collapse into six upstream mechanisms — and a mechanism can be fixed once and stopped from recurring. Each one below is admissible only because something in the data proves it.
An orientation set per feed, set wrong on the feeds the retailer owns
The tell: the reversal follows brand ownership almost perfectly — near-total on the owned brands, rare on the resold ones. Data entry does not sort itself by owner. It produced 374 reversed listings.
A size's dimension set attached to the wrong variant
The tell: every published value is a valid mattress size, just not the one in the name. It produced a series in which no listing is correct, a half-bed listing measured as the whole set, and split listings carrying a California King's footprint.
One variant group dropping fields in transit
The tell: a single brand-and-size group is blank on a field its siblings all carry, while everything else on it is present. It produced fifteen split California Kings with no height or depth, and listings with no price or availability.
Each product published at two addresses
The tell: the two records are identical on every field, and both are in the sitemap. It produced forty-one products counted twice.
The specification living only in the name
The tell: size, firmness and construction have no fields of their own, so wherever a name is shortened the specification goes with it. It produced split names that lost their series, a series with no size in its names, and no size filter that can trust its source.
A default or an internal value standing in for a real one
The tell: the same value on every record, or a value meant for staff reaching the storefront. It produced a flat sitemap priority, dead pages dated like live ones, and an internal label and a clearance label serving as mattress categories.
Correcting 374 reversed listings by hand corrects 374 listings. Correcting one orientation per feed corrects most of them at once — and every mattress those feeds deliver next.
38%
In a September 2025 survey of 1,000 U.S. consumers, 38% said they will abandon a site if they cannot easily find key features such as sizing and materials — and a further 39% said they will go find that information on a competitor's site instead. On a mattress, sizing is not one key feature among several. It is the first decision, and it is the field this study found reversed or misplaced.
Akeneo PX Pulse survey, conducted by Dynata, September 2025, n=1,000 U.S. consumers aged 18+
A retailer that owns its own brands has a structural advantage in mattresses: it controls the product, the margin and the story. This catalog shows the other side of that advantage. The owned brands are the ones the record gets wrong, and the national brands are almost all correct — a split that points at the path owned-brand data takes to the storefront, not at the people entering it.
What this class of defect costs is not a lost sale so much as the wrong one. Anything that reads dimensions to decide what is a queen — a comparison engine, a marketplace feed, an assistant — reads a queen turned sideways on every owned-brand listing. A customer choosing a California King from one series is shown a Queen's measurements. And a mattress chosen on dimensions that belong to another bed comes back: delivered, unwrapped and returned.
A second clock runs underneath. The shopper comparing mattresses increasingly starts with an assistant rather than a search box, and an assistant does not look at a showroom tag. It reads the fields — matches a size across retailers, checks a thickness, confirms a price. On this catalog it finds no size field, dimensions turned sideways on the owned brands, listings it cannot price and products it sees twice. That is not a ranking to trade off. It is an absence, and absences do not improve on their own.
Three moves, in order. The first touches the most listings and needs the least new information.
This pass read and diagnosed. The same structural understanding powers what follows — turning a diagnosed catalog into one a shopper can choose a size from with confidence.
1 · Apply the corrections the record already answers for itself
Correct the orientation on the owned-brand feeds toward the convention the national brands already follow. Restore height and depth on the split California Kings from their own names and their size. Structure size, firmness and construction into fields, cross-checked so a name and its numbers must agree before a value is written. Fill blank brands from product names and return one-off categories to the vocabulary their siblings use. Approved as patterns, not product by product.
2 · Settle the questions that gate the rest
Whether the second, master-style pages are products or addresses. Whether split listings describe one half or the pair. Whether the listings with no price are discontinued or broken, and whether the hidden kit record should be public at all. Those answers come from inside the business before the remaining corrections can be applied.
3 · Hold the line at intake
Keep ongoing intelligence where each brand's feed meets the storefront, so the next feed that arrives the wrong way round — including the next brand the retailer adds — is caught as it lands rather than found later by a customer choosing a bed on it.
This is how EKOM moves a catalog from insight to impact —
and keeps every size reading right as the business grows.