Resolution Layer Case Study · Read & Diagnose
A luxury appliance dealer.
A catalog read across built-in refrigeration, wall ovens and dishwashers — for a dealer that authors almost none of its own product content, and where the defects go instead.
This is a real EKOM catalog analysis, with the dealer's identity removed. The client is an independent luxury appliance and design dealer — Sub-Zero, Wolf, Thermador, ASKO, Perlick, Hoshizaki and the Whirlpool family of brands — that has moved the weight of its business online. EKOM read 2,646 products across three complete appliance categories and an entire furniture and miscellaneous section, and surfaced 31 flagged items, 30 of which are actionable catalog defects.
What makes this catalog unusual is what is right about it. The dealer authors almost none of its own product content — every description and image arrives from an upstream manufacturer feed. So there are very few copy defects, because there is very little copy of theirs to be wrong. The defect count is low for a catalog this size, and that is itself the finding.
Where a dealer's content comes from its manufacturers, the defects do not live in the content. They live in the thin layer the dealer adds on top — the category, the brand tag, the choice of photograph — because that layer is the only part of the record with nothing checking it.
Third-party brand and model names are kept exactly as they appeared, which on a dealer catalog is nearly every proper noun in this document. Only the dealer's own identity, location and history have been removed.
What's inside
- The findings a shopper hits first — a drafting note shipped to customers, and a fifth of the appliance catalog showing another product's photograph.
- A record that disagrees with itself — the receipt device that needs no outside data to prove a defect.
- Where it traces back — a four-level category path with one word in all four slots, and three specs that exist only as prose.
- The machine-readable surface — what a search engine and an AI assistant can actually read off these pages.
- What it means, and what's next — from diagnosis to a catalog a shopper can filter.
Thirty actionable findings across twenty defect themes. Severity bands are the analysis engine's own; the theme grouping is EKOM's.
30
Actionable findings
(31 flagged)
Where the thirty concentrate
| Group | Themes | Findings | Who feels it |
|---|
| Identity and classification — wrong department, wrong brand tag, wrong product class | 8 | 10 | Shoppers + Ops |
| Media — distinct products resolving to one photograph | 1 | 5 | Shoppers |
| Values and vocabulary — a published note, a self-contradicting record, a wrong finish | 6 | 6 | Shoppers + Ops |
| Completeness — empty descriptions, boilerplate, missing prices, missing spec fields | 4 | 7 | Shoppers |
| Duplication — one manufacturer part listed under several sibling brands | 1 | 2 | Ops |
Read this as triage, not as a verified defect list. These are machine-surfaced signals from a cold read of published data. A small share may turn out to be intentional — a manufacturer pricing policy, a deliberate merchandising exception. Where a finding needs a business decision rather than a correction, the analysis says so, and one of the thirty is flagged for human review rather than asserted.
What a shopper runs into first.
Highest impact — what a customer meets before anyone inside the business does.
Lead finding · published internal note
An editorial to-do, shipped verbatim to customers.
The live description on a Thermador Star Sapphire dishwasher — model DWHD661EFM, a $2,499 appliance — ends mid-thought with a drafting note that was never taken out: …Energy Star Certified to help you save water and energy. We should list DBA & Energy Star. One record, and the cheapest fix in the report — the correction is a deletion of the final sentence. It is the clearest illustration of the underlying condition: nothing sits between what someone types into the record and what a customer reads.
A fifth of the appliance catalog shows another product's photograph
22% of the appliances examined share their primary photograph with a different product of the same brand. Not color variants of one item — separate models, separate features, separate prices. The Whirlpool dishwashers AI Intelligent Wash, Soil Sensor and Spin&Load Rack resolve to one photograph; so do the KitchenAid Third Level Utensil Rack, 360° Max Jets™ and Third Level Jet Rack. A shopper putting either set side by side sees the same picture three times.
Two cases matter more than the rest, because the photograph is the only place the difference is visible at all. The Sub-Zero PRO3650LH and PRO3650RH are the left-hinge and right-hinge builds of the same refrigerator, on one image that cannot depict both door swings. The wall ovens WOED7027PZ at 8.6 cu. ft. and WOED7030PZ at 10.0 cu. ft. share one image, so the larger oven is pictured as the smaller one. The pattern runs outside appliances too, through Perlick's C-Series beer dispensers — single-tap and dual-tap, hinge left and hinge right, stainless and panel-ready, all on a single photograph.
Counted conservatively. The 22% counts same-brand cases only. Products sharing an image across different brands were excluded — an identical part genuinely sold under several Whirlpool-family labels is expected, not defective — and removing them cut the raw figure by roughly a fifth.
A record that disagrees with itself.
The most persuasive kind of finding, because nothing external is needed to prove it — and the reason complete categories were read rather than a scatter sample.
Thermador PO302W does not agree with itself about how many cavities it has. Its own sibling settles the argument.
Field
Value
Says
Title
contains double oven
Double
Category
double-wall-electric-oven
Double
Description
this 30-inch Thermador single oven
Single
Capacity, in description
4.5 cu. ft., one cavity
Single
Sibling POD302W, for comparison
Double Oven pairs two True Convection Ovens.
Double
A double wall oven and a single wall oven are different cabinet cutouts. Whichever half of this record is wrong, the customer who acts on the wrong half has a cabinetry problem, not a returns problem.
The same shape, six more times
| What the record says | What the record also says |
|---|
| Category built-in-dishwasher | Description: stainless free standing/outdoor XXL dishwasher. (ASKO DOD561TXXLS) |
| Category single-wall-electric-oven | Title: 24 in. Under the Counter Convection Microwave Drawer™ Oven Pedestal (Sharp SKCD24U0GS) |
| Category no-freezer-built-in-refrigerator | Title: 36" Classic Refrigerator with Glass Door — and the records either side of it carry glass-door-built-in-refrigerator |
| Brand MAYTAG | Description: KitchenAid panel Ready handle features red medalions and fits dishwsher models KDTF324PPA & KDTF924PPA |
| Color Bisque — a beige finish | Title: Touch-Up Paint with a Brush - Milkshake 0.6-oz. Milkshake is a white finish; the Whirlpool sibling on the same model carries White |
| Category dishwasher-accessories | A cast-aluminum griddle for gas ranges, a laundry detergent dispenser, and a refrigerator light bulb |
None of these need outside data to catch. In every case the contradiction is sitting in two fields of the same record, or in the record either side of it — which is why complete categories were read rather than a scatter sample. Split a variant family apart and the comparison becomes impossible.
Two structural conditions sit upstream of most of what surfaced.
A catch-all standing in for a taxonomy
Just over half of the furniture and miscellaneous section carries the same value at every level of its four-level category path: furniture / miscellaneous-furniture / miscellaneous-furniture / miscellaneous-furniture. Filed under that one node: commercial heated cabinets, beer dispensers, mixer attachments, fire tables, air purifiers and televisions. Meaningful taxonomy nodes exist in this catalog — they are used on a minority of items. Where a path is four levels deep and all four say the same word, there is no faceted navigation to build and no way for a shopper to arrive anywhere except by search.
The section is also not furniture. Its contents include Refrigerator Herb Tender, 27-IN Warming Drawer Heat Deflector and Drip Tray for Smart Oven — appliance accessories, in a department named for something else. EKOM's profiling pass, which classifies a section on its contents rather than its label, read this one as appliances before any analysis ran. That misclassification is not a bug in the read. It is the finding, arrived at independently.
The three specs an appliance buyer filters on do not exist as fields
This catalog carries no discrete specification attributes. Width, capacity and noise level are all present — every one of them — but they are present inside titles and descriptions, as prose. Widths from 18 to 60 inches, sitting in titles. 16 place settings and 4.5 cu. ft., sitting in descriptions. Noise levels from 37 to 60 dBA, likewise.
For a built-in appliance, width decides whether the product fits the hole in the cabinetry — the one specification a customer cannot get wrong without sending the unit back. Capacity and noise level are what a dishwasher shopper compares after price. None of the three can be sorted, filtered or faceted on today, because prose is not a value. All three are extractable from text that already exists.
And a third condition, outside the catalog
43% of the appliances examined carry no validated price. Some of that is very likely deliberate — Sub-Zero and Wolf both operate minimum-advertised-price policies, and their authorized dealers routinely publish a quote request rather than a number. The question is not whether the price is missing. It is whether the catalog can tell the difference between a price it does not have and a price it is not allowed to publish. A filter cannot, and the shopper narrowing by budget loses both.
The machine-readable surface.
Measured directly across 2,609 product pages. A separate question from whether the catalog data is right — and on this catalog, a starker one.
| What a search engine or AI assistant finds | Coverage |
|---|
| Canonical URL | 2,609 / 2,609 |
| Meta description — present, and every one unique | 2,609 / 2,609 |
| Open Graph tags | 2,609 / 2,609 |
| Structured product data — Product, Offer, Brand | 0 / 2,609 |
A unique meta description on every product page across a catalog this size is uncommon, and worth saying before the rest of this page. That is not the default.
The answer layer has nothing to read
No structured product data on any page measured.
No Product markup, no Offer, no Brand, no rating. Google rich results have nothing to extract, and neither does an assistant answering “which built-in dishwasher should I buy.” A dealer that holds the inventory, the brand relationships and the expertise cannot be cited as the source for the question it is best qualified to answer. Two further gaps compound it: each product page arrives carrying exactly two images, both the site logo — the product photographs appear only once the page's own code runs, which Google does and many answer engines do not — and once they do render, their alt text is positional boilerplate, Thumbnail for product image 1.
| Page-level defect | Products | Share |
|---|
| Page title longer than a search result will display | 2,506 | 96% |
| Meta description longer than a search result will display | 2,405 | 92% |
| Page title shared with another product | 176 | 7% |
| Meta description too short to be useful | 46 | 2% |
The truncation figures are a template problem, not a per-product one — the title pattern is brand, product name, model, business name, which overflows on almost everything. The 176 duplicate titles are the per-product defect worth chasing: two distinct products competing for one search result.
How EKOM reads this catalog.
Why complete categories, read against themselves, catch what a spot-check misses.
1
Read complete categories
Three whole appliance categories and an entire furniture and miscellaneous section, normalized to one record per product — no file handoff, no credentials, no integration.
Complete categories, not a scatter sample, so a product can be compared against its own sibling variants.
→
2
Profile on contents
Map the shape of every field — what it holds, how consistently, where conventions break — and classify each section on what is in it rather than what it is called.
The section named furniture came back classified as appliances. The catch-all finding surfaced here, before anyone went looking.
→
3
Analyze and reconcile
Analytical passes across both sections, reconciled into typed, deduplicated findings — each with a severity, an intended fix, and the exact records it applies to.
30 actionable findings, 9 of them carrying a corrected value derivable from the record itself.
A completeness check would have passed almost all of this catalog. Every field in every finding above is populated — the category is filled in, the brand is filled in, the image is there. What a validator cannot ask is whether the value is right, and on a dealer catalog that is the only question worth asking, because correctness is the one thing the manufacturer feed does not supply.
About this analysis
This is EKOM's first read of this dealer's catalog. The engine read complete product categories rather than a random sample, so each product could be compared against its own sibling variants — which is what makes a self-contradicting record detectable at all. The findings here are that read, with the dealer's identity, location and history removed and third-party brand and model names preserved. Coverage is a scoped part of the catalog, and nothing here speaks to the categories that were not examined.
as high as90%
Baymard Institute benchmarked the 344 top-grossing US and European e-commerce sites and found that sites with mediocre product-list usability saw abandonment rates of 67–90%, against 17–33% for sites with even a slightly optimized filtering toolset. It also found 36% of those sites carried flaws severe enough to be actively harmful to a shopper's ability to find and select a product. Filtering is the whole of that finding — and on this catalog a shopper cannot filter by noise level, a builder cannot filter by width, and half a department resolves to a single node with nothing to browse.
Baymard Institute, E-Commerce Product Lists & Filtering UX benchmark
A dealer whose product content is manufacturer-supplied starts from an advantage most merchants do not have: the copy is internally consistent by construction, and the defect count is correspondingly low. That advantage is real, and it is exactly why the remaining defects are so easy to miss. They are not distributed across the catalog like typos. They are concentrated in one thin, hand-maintained layer — the category, the brand tag, the photograph — and everything in that layer arrives unchecked.
The consequences are not cosmetic. A wall oven whose record cannot agree on one cavity or two is a cabinet cutout, not a return. Touch-up paint sold as bisque and shipped as milkshake is an appliance repainted the wrong shade. A dishwasher pictured as a different dishwasher converts a comparison into a return, or into nothing at all. And a specification that lives inside a sentence cannot be filtered on, which means the shopper who knows exactly what they want is the one most likely to leave.
Three moves, in order. The first needs no new data at all.
This pass read and diagnosed. The same structural understanding powers the work that follows — turning a diagnosed catalog into one a shopper can actually navigate.
1 · Apply the corrections that are already derivable
Nine of the thirty findings carry a corrected value the record itself supplies — the published note, the two brand tags, the wrong finish, the separated variant, the split product family, and the extraction of width, capacity and noise level out of existing text. Approved as patterns, not product by product.
2 · Give the catalog a taxonomy and a spec layer
Replace the four-level catch-all with real nodes, and stand up the filterable fields a built-in appliance buyer needs — so that width, capacity and noise level stop being sentences and start being facets.
3 · Hold the line at intake
Keep ongoing intelligence where the manufacturer feed meets the dealer's own layer, so the next category assignment, brand tag and image choice is checked as it lands rather than discovered later by a customer.
This is how EKOM moves a catalog from insight to impact —
and keeps every listing reading right as the business grows.