Resolution Layer Case Study · Read & Diagnose
Refrigerator product pages.
A catalog read across the refrigerator pages of a national home improvement retailer — where a shopper sizes a kitchen by capacity and width, and a search engine matches each listing by its model number and UPC.
This is a real EKOM catalog analysis, with the client's identity removed. The subject is a national home improvement retailer that sells refrigerators from several major brands, each model with its own page. EKOM read a random sample of those pages the way a shopper does, then held each page against itself: its title against its web address, its address against its specifications, and what it shows a shopper against what it hands a search engine. Ten findings are reported — two high, three medium and five low — along with two context items that are not rated.
What makes this catalog different is that the pages mostly agree with themselves. Every page read shows its model number, and the capacity in the title matches the specification table wherever a table lists one. That consistency is what makes the exceptions easy to see, and it is also what makes a gap in the data behind the page stand out: the model number is on the page, and it is not in the data a search engine reads.
The shopper reads the page. The search engine reads the data behind it. When that data leaves out what the page shows, the page can look right and the listing can still be unmatched to the same product elsewhere and unaware of whether the item can be bought.
Product names, models and values are described rather than quoted here. The counts are exact counts of pages in a sample, and they describe that sample; they should not be scaled to a catalog.
What's inside
- At a glance — the severity split, where the ten concentrate, and how the read was done.
- What a search engine is given — the lead finding, and a page with no product record at all.
- The same page, different facts — capacity and ENERGY STAR status that disagree between a page's address and its title or specifications.
- Width, smaller items, and what is strong — what is rated low, and what already works.
- What a page says while it sells — two context items, not rated.
- What is actionable, and what is held — the corrections that apply as patterns, and the questions only the retailer can answer.
Nothing was supplied and no internal system was touched. Everything below was read from the retailer's public refrigerator pages, at a shopper's pace, following the robots.txt and sitemaps the retailer publishes.
Where the ten concentrate
Theme
What it is
Findings
Product record
The data a page gives search engines leaves out the UPC, model number and stock status, and one page gives no product record at all
2
Capacity
The capacity in a web address disagrees with the capacity in the title
1
ENERGY STAR
An address says ENERGY STAR where the specifications say the refrigerator is not certified
1
Width
A title width the actual width does not support
1
Smaller items
Raw field names, an address that describes a different product, a capacity field holding the whole-unit figure, a model number that differs by a suffix, and title typography
5
How EKOM read this catalog
1
Read
Each page as a shopper sees it, and the product data the page embeds for search engines, copied as shown rather than retyped.
A random sample of refrigerator pages from several major brands.
→
2
Compare
Test the title against the web address and the specifications: capacity, width, ENERGY STAR status, color and model number.
Where a page agrees with itself, that is reported too.
→
3
Check
Set the product data a page gives search engines beside what the page shows a shopper: model, UPC, price and stock.
Gaps, mismatches and offers that outlive stock.
A sample, not a census, and leads rather than verified defects. Nothing here has been confirmed with the retailer, and some values may originate upstream of it; a shopper reads them on the retailer's page either way. Point-in-time. Pages were read on one day and a few headline items were re-checked two days later. Storefronts move, prices are store-specific, and a finding may have been corrected since. Severity ratings are EKOM's.
What a search engine is given.
Each refrigerator page embeds product data for search engines. A shopper sees a model number and, where it applies, an out-of-stock notice. The data a search engine reads leaves out both, along with the UPC.
High · Structured data
No page read gives search engines a UPC, a model number or a stock status, though the page itself carries all three.
Every page shows its model number to shoppers, the product record behind the page carries the UPC, and several pages tell the shopper the item is out of stock. The structured data on every page that carries a product record holds a name, a SKU and a price, and no GTIN, no model number (MPN) and no availability. A search engine cannot match these listings to the same product elsewhere, and cannot tell which of them are unavailable. The fields already exist on the page, so filling the product record is a mapping job, not new data collection.
Field
What a shopper sees / what the structured data holds
Verdict
Model number
shown on the page / no model number (MPN)
missing
UPC
in the product record behind the page / no GTIN
missing
Stock
an out-of-stock notice on some pages / no availability
missing
Name, SKU, price
shown / present
present
Medium · Product record
One page gives search engines no product record at all.
A French door model embeds breadcrumbs, images and videos but no product record, so a search engine reading it finds no name, price or identifier. No other page read lacks one. The other pages show what the platform emits when it works, so finding why this one did not get the same record is a one-page diagnosis, and the fix then applies to any page built the same way.
The same page, different facts.
A refrigerator page names its capacity in the title, in the specifications and in the web address. These should say the same thing. On some pages they do not.
High · Capacity
On eleven pages the capacity in the web address disagrees with the title; on two the gap is large.
On a built-in French door model the address gives a figure several cubic feet below the title and specifications, and it appears to be the fresh-food section alone. On a bottom freezer the gap is two to three cubic feet. Nine more differ by half a cubic foot or less, a rounding or a fridge-plus-freezer figure against the rated total, and may be harmless. A web address is read by search engines and shown when a link is shared. Where it and the title disagree, a shopper and a system get different answers to the first question asked about a refrigerator: how big is it?
Medium · ENERGY STAR
Two web addresses say ENERGY STAR for refrigerators whose specifications say they are not certified.
Both specifications list ENERGY STAR Certified: No, and neither title claims certification. One of the two addresses also gives a capacity a fraction of a cubic foot below its title. ENERGY STAR is a status shoppers filter by and rebate programs check, so the address and the specification need to agree. Which one is right is a question for whoever owns certification, not for the data.
A standing check that compares the capacity in all three places, and any ENERGY STAR mention with the certification field, would catch each wherever it recurs, and would raise the page for review rather than change it.
Width, smaller items, and what is strong.
Medium · Width
A bottom freezer is titled 33 inches wide and lists an actual width more than three inches less.
Its size class and its common width agree with the title. Its actual width reads the same figure the page gives for its depth with the door removed. Every other page that lists both a title width and an actual width is within three quarters of an inch. A shopper measuring an opening gets two answers from one page, and a check of actual width against title width would raise it.
Five further items are leads rather than defects. Each is a value that sits outside the pattern its neighbors set, and each may turn out to be legitimate, which is why they are rated low.
Low · A new listing's field names
A recently listed chest freezer sends raw field names, where the other pages send readable labels, and a certification number of TBC. It also has no breadcrumb trail.
Low · An address that describes the product differently
On one page the address names a different finish from the title and the color field, and the model in that finish is a separate listing. On another the address says the product converts from freezer to refrigerator, which its title and specifications do not mention.
Low · Refrigerator capacity
One page lists an overall capacity, a freezer capacity and a refrigerator capacity, and the refrigerator figure is the whole-unit figure.
Low · Model number
One page shows a model number on the page and a longer one, with an extra two-letter suffix, in its specifications.
Low · Title typography
Seven titles carry an unmatched closing parenthesis, ten write “3 -Door” or “4 -Door” with a space before the hyphen, and one series name is spelled with a stray capital in its title, address and specifications.
What is strong. These are pages that mostly agree with themselves, which is why the exceptions are easy to see. Every page read shows its model number and item number. The capacity in the title agrees with the specification table on every page that lists one, as does the ENERGY STAR status, and the title width is within three quarters of an inch of the actual width everywhere except the one page above. A rule written against pages this consistent is practical rather than a line-by-line rewrite.
What a page says while it sells.
Everything above is about what a page states. This page is about two things a page does at the moment of sale, neither of which is rated here. They are context worth knowing, because each changes what a shopper or a search engine concludes from the same page.
Not rated · Out-of-stock pages still show offers
At least seven of the pages that say the item is out of stock still show the monthly-payment offer beside the price, and three of the out-of-stock pages also show a sale price. A payment offer on an item that cannot be bought is a small contradiction, and the shopper meets it at the worst moment. A display rule that ties offers and sale prices to stock status removes it, and it needs no change to the product data itself.
Not rated · The price a search engine reads
On two pages the price a search engine reads is not the price in the page's own product record. On one, the page and its structured data show an advertised price with a prompt to view a lower price in the cart, while the record behind the page holds a price well below it. On an out-of-stock page, the record holds a sale price that neither the page nor its structured data shows. Showing an advertised price above the cart price is a normal arrangement with a manufacturer, and this analysis does not treat it as a gap. The consequence is narrow: a search engine reads the advertised price, not the one the shopper pays. Whether the structured data should carry the advertised price, the cart price or both is the retailer's call.
A diagnosis is not a deliverable. Each correction below is a pattern applied to every page that matches it, not a list of individual edits. The retailer approves the pattern, not the individual pages.
Complete the product record on every page
The model number, the UPC and the stock status are already on the page or in the product record behind it. Mapping them into the structured data, and giving the page that has none a record, is one change applied to every page built the same way. A search engine can then match each listing to the same product elsewhere, and can tell which are unavailable.
Hold the address, title and specifications to one capacity
A standing check compares the capacity in all three places and raises any page where they differ by more than rounding. For the pages found, the retailer's own product record settles which figure is right.
Tie every ENERGY STAR mention to the certification field
Any ENERGY STAR mention in an address, title or description must match the certification field, and a page where they disagree is raised for review, not changed. Once the status of the two pages found is confirmed, the rule applies everywhere.
Add plausibility checks for width and capacity fields
Actual width is checked against the width in the title, and refrigerator and freezer capacity against the overall figure. Pages outside the pattern are raised for review, not changed.
Tie offers and sale prices to stock status
A page that says an item is out of stock stops showing financing offers and sale prices beside it, and shows them again when the item returns.
Normalize titles and addresses
Parentheses are balanced, “3-Door” is written one way, and a series name is spelled once, in the title, the address and the specifications at the same time. The same rule catches any new title with the same break.
Held for the retailer's decision.
These questions block correction work. The pages show the conflict but cannot resolve it, and the right answer exists inside the retailer. None should be corrected automatically, and this analysis does not pretend otherwise.
Question
What the pages show
Who answers it
Which capacity is right when the address, title and specifications disagree
Two pages with a large gap, and nine with a small one.
Merchandising and product data.
Whether a refrigerator is ENERGY STAR certified
Two addresses say ENERGY STAR; both specifications say not certified.
Whoever owns certification, and the team that receives manufacturer data.
What a refrigerator's actual width is
A bottom freezer titled 33 inches with a smaller actual width.
The team that owns the manufacturer specification feed.
Whether the product record is meant to carry the UPC, model number and stock
None of the three in the structured data of any page read.
The e-commerce and SEO owners.
Whether offers should stay on pages that are out of stock
At least seven out-of-stock pages show monthly-payment offers, and three also show a sale price.
E-commerce and merchandising.
Five questions gate the rest of the work. Only the retailer's people can say which of two conflicting figures is right, which certification status stands, and what the product record is meant to give search engines. Everything downstream of those answers — the normalizing, the checking of each page against itself, the enrichment, and the watching that catches the next mismatched address — is EKOM's.
What this means, and what's next.
A home improvement retailer has told its customers it wants them better informed. Online, that rides on the page: the capacity that says whether a refrigerator fills the space, the width that says whether it fits, the ENERGY STAR status behind a filter or a rebate, and the model number that lets a search engine, an assistant or a shopper comparing stores know it is the same product.
The costs are concrete. A shopper sizing a kitchen can read one refrigerator at one capacity in the title and another in the address. A search engine matching listings across sellers has no UPC or model number to match on. A page that says an item is out of stock still invites a shopper to finance it. And none of it needed access to the retailer's systems: it was read from the retailer's public pages.
This pass read and diagnosed. What EKOM brings is not the ability to see what is wrong, but the commitment to stay in it until it is right, and to keep it right as the catalog grows.
1 · Apply the corrections the pages already answer for themselves
Fill the product record from fields the page already holds, normalize the titles and addresses, and add the plausibility checks. Approved as patterns, not product by product.
2 · Settle the questions that gate the rest
Which capacity stands, which certification status stands, what the product record is meant to publish, and whether offers stay on an out-of-stock page. Each is a judgment the data can show but cannot make.
3 · Hold the line as the catalog grows
Keep a standing check on the pages, so a mismatched address, a missing record or an offer on an unavailable item is caught if it comes back as the catalog grows.
This is how EKOM moves a catalog from insight to impact —
and keeps the machine's copy of every product agreeing with the page as the business grows.