Resolution Layer Case Study · Read & Diagnose
Handbag colorways.
A catalog read across a direct-to-consumer handbag brand's complete public catalog — where every colorway is published as its own product, and most of them share a name.
This is a real EKOM catalog analysis, with the client's identity removed. The subject is a direct-to-consumer handbag brand that sells in drops: each colorway of a bag — a color, a material, a hardware finish — is published as its own product, with its own page and its own address. EKOM read the whole public catalog and every product page. Thirty-five findings are reported here. Page eight accounts for the ones that were not.
In a drop business the colorway is the product. A shopper does not want the bag; she wants the bag in that color, in that finish. So the catalog has to carry the colorway in the places that name a product — the title, the page title a search engine indexes, the data a shopping feed reads. This read asks whether a shopper, a marketplace or an AI assistant, reading the record alone, could tell which colorway it is looking at.
When every colorway is its own product, the name is the only thing that tells two listings apart. A catalog that names the family and hides the color in an option has published the same product dozens of times.
Product names, addresses and values are described rather than quoted here. Counts are exact, as of the day the catalog was read.
What's inside
- At a glance — the severity split, where the thirty-five concentrate, and how the read was done.
- Many colorways, one name — the lead finding, in the catalog and in the page titles Google indexes.
- What Google, feeds and assistants are told — search snippets that were never written, and the brand split by spelling.
- Listings that describe another bag — copy and addresses carried over from the colorway before.
- Checkout tools, types and the rest — non-products in the feed, a product type that names the family, and the smaller items.
- Confirming against source, and where it traces back — what was withdrawn or held, and the six mechanisms EKOM reads in the rest.
Nothing was supplied and no internal system was touched. Everything below was read from the product feed the brand publishes, and from its product pages as served.
344
Products sharing
one of 42 titles
284
Colorway pages with no
color in the search title
59
Product types — mostly
families and size runs
6
Likely mechanisms, most
repeated with every release
Where the thirty-five concentrate
Theme
What it is
Findings
Copy that describes a different product
Descriptions, addresses and tags naming another colorway
4
Non-products in the catalog
Checkout add-ons published beside the bags
1
What Google and AI assistants read
Repeated page titles, unwritten snippets, the brand split by spelling
6
Categories and names
Repeated titles, a product type that names the family, the brand spelled two ways
5
Specifications and prices
Dimensions, a strap length, SKUs, options, a steep markdown
9
Tags
Junk, test, stale and contradictory tags
5
Images and description copy
Screenshots, other products' files, code left in descriptions
5
How EKOM read this catalog
1
Read
The brand's public product feed, every product and variant, and every product page as served. No file handoff, no credentials, no integration.
The whole catalog, not a sample.
→
2
Analyze
Infer what each field is for from how the catalog itself uses it, then hold every product to that. Surface the disagreements, rate severity, and tie each to the products behind it.
35 findings, 7 themes, 6 mechanisms.
→
3
Check
An analyst checked every figure against the feed, and re-opened a sample of the headline findings on the live store before writing.
One finding withdrawn, one held, three set aside as observations.
Signals from a read of the published data, checked by an analyst; not a verified defect list. Some will prove intentional — a tag that drives a collection, a price set for a promotion — which is why each is tied to the values behind it. One catalog, one day. The data is a point-in-time read, and the catalog has changed since.
Many colorways, one name.
Each colorway is its own product, but most share their family's name.
High · Categories and names
344 products share one of just 42 titles. The color lives only in an option value.
The brand's signature bag appears as fifty-four products with one identical title, and its smaller version as fifty more. Each is a different color, material or hardware finish. In a list of results, a feed or a comparison they are the same product repeated, because the difference sits in the color choice on the product — a field most channels treat as a variant detail, not as the product's name. One bag family is titled four different ways, one of them with a double space.
High · What Google and AI assistants read
The page titles Google indexes repeat the same way.
Each page title is the product title followed by the store name, so the repetition carries straight through: 344 pages share one of 42 page titles. On 284 pages for products sold in a color, the color is not in the page title at all.
A shopper searching for one colorway is shown dozens of results with the same name. A search engine sees dozens of pages competing for one title and has to guess which to rank. And an assistant asked for a particular color of the signature bag finds no color in the title to match. The color exists in the record — in the option value — but that is not where a name is read.
What Google, shopping feeds and AI assistants are told.
Every product page carries a layer a shopper never sees: product data embedded for Google, shopping feeds and AI assistants, and the description shown under a search result. Price, availability and SKU in that data match the feed on every page, and every page points to itself as its main address. The rest needs work.
High · Search snippets
The descriptions shown in search results are not written. They are cut off.
The median meta description is exactly 320 characters — the platform's automatic cut from the product description, with line breaks removed so sentences run together. 325 run past the roughly 160 characters a search result shows, 356 open with the same template phrase, and a set of twelve phone cases share one identical description.
High · One brand, two spellings
The brand appears under two spellings in the product data, plus a sub-brand.
The brand each page declares follows the vendor field: the full name on most products and an abbreviated form on 13. Google Shopping, marketplaces and AI assistants can treat a different spelling as a different brand. A separate apparel label on 48 may be a deliberate sub-brand — the brand's call. A handful of addresses also misspell the product's own name, one checkout page is open to Google with no product data at all, and the barcode is kept as description text rather than in its field.
Listings that describe another bag.
When the color is the product, a listing that names the wrong color is describing a different product. Four of the five critical findings are here; the fifth is on the next page.
Critical · Copy that describes a different product
Two bag charms, one green and one silver, are both described as the orange charm.
Both descriptions read as the orange charm's copy. The silver one also promises gold hardware, while its own title and option say silver.
Critical · Copy that describes a different product
Six product addresses name a color, material or style the product is not.
An address that names the wrong skin color; a black bag that is green; the wrong hardware finish; a clutch that is the signature bag; one material named for another; one green named for another green. In each case the product's own SKU or option says otherwise.
Critical · Copy that describes a different product
Fourteen addresses read "copy-of" or "-copy", and some listings contradict themselves.
The copy addresses are leftovers from duplicating a listing to make the next colorway. Elsewhere, a backpack is tagged silver hardware and described as gold, and two lace bags say white satin in title and option while their addresses and SKUs say nude — as does one of their descriptions.
The pattern is consistent with one habit: a new colorway made by copying the last one, with not every field rewritten. That is a fast way to launch a drop, and it carries forward whatever the copied listing said.
Checkout tools, types and the rest.
Critical · Non-products in the catalog
Shipping protection, a return-coverage fee and a gift-card app are published as products beside the bags.
Each is a legitimate checkout add-on. In the public feed — the one shopping engines, marketplaces and AI agents read — they look like products: the shipping protection carries 76 price variants from under a dollar to over $140, the return fee has a SKU no product would carry, and the gift card has twelve values and no SKUs.
High · Categories and names
The product type holds 59 values — mostly bag families, size runs and blanks rather than categories.
Three sizes of one bag are three separate types. Every shoe is typed by a size range that is not the US sizes it is sold in, so a category built on this field files the shoes under a size, not under footwear. One type is misspelled on five products, and three products have no type.
The smaller items
- Specifications and prices (9): dimensions that look copied between models; a shoulder strap listed ten inches short; SKU typos and mixed formats; option values misspelled; a phone model stored as a size; the size option named four ways; a steep markdown on one phone case; and the barcode kept only as text.
- Tags (5): a phone-number-like junk tag on 102 products; test tags on live products; a campaign tag from a past season on every product; typos in collection tags; and contradictions — a sold-out bag tagged available, raffia bags tagged as clutches.
- Images and description copy (5): eleven images that are screenshots or AI-generated files; other products' image files reused; editor code left in descriptions (stray meta tags alone in 189); a shoe description that links to another store; and four listings with a one-line description.
Confirming against source.
The automated analysis raised thirty-three findings, and EKOM's analyst added seven from a read of every product page. Thirty-five are reported. Every one of the others is accounted for here.
Withdrawn · one finding, and part of another
One finding described a word left at the end of descriptions; it did not hold on the store's own descriptions, and was withdrawn. Part of the barcode finding was withdrawn too: barcodes written with leading zeros are a valid long form of the same UPC, not an error. The rest of that finding stands.
Observations, not defects · three findings
A uniform 35% markdown on raffia and linen bags reads as a deliberate campaign. Hardware finish has no field of its own. And whether the apparel label is a deliberate sub-brand is for the brand to say. None is counted.
Held after a live check · one finding
The page read suggested each product page's first heading was a cart message. Re-opened live, that heading sits inside a hidden pop-up template, which search engines do not read as a heading. It is not reported.
Re-checked live
Before writing, a sample of the headline findings was re-opened on the live store. The copy carried over to the silver charm, two of the addresses that name another colorway, the indexable checkout page and the missing barcode all still stood. One of the two charms had left the public feed.
Holding a finding back is part of the read. A report that had shipped the heading finding would have told a brand its pages were wrong where the pages were right.
Thirty-five findings sound like thirty-five problems. Most of them collapse into six likely mechanisms — and a mechanism can be corrected once and stopped from recurring. Each tell below is a pattern in the data; the mechanism is EKOM's reading of it.
Each colorway published as its own product, named only for its family
The tell: the title is the family name and the color is written only into the option, so every new colorway adds a listing with the same name.
A new colorway made by copying the last one
The tell: fourteen addresses that still say "copy", and wrong values that read as other listings' real values — carried over, not mistyped.
Facts kept as free text where a field should hold them
The tell: the barcode in a description, a phone model under "Size", hardware finish only in titles and options, a product type used for the family.
Checkout apps writing into the product catalog
The tell: entries that carry an app's name as vendor, with prices or SKUs no product would have.
Platform defaults never overwritten
The tell: snippets cut at a median of exactly 320 characters, and page titles that are the product title plus the store name.
Campaign, test and pasted-in residue never cleared
The tell: a past season's tag on every product, test tags on live products, and editor code inside descriptions.
What this means, and what's next.
A drop business depends on shoppers who know exactly which colorway they want. The catalog is where that knowledge gets lost: a shopper looking for one colorway meets dozens of listings with the same name, a search engine meets dozens of pages with the same title, and an assistant reading the fields finds a family name instead of a product name, a product type that files shoes under a size range, the brand under two spellings, and a shipping-protection policy listed beside the bags.
None of this needed access to the brand's systems. Every defect here was found by reading data the store already publishes to anyone who asks — so the same read is open to any search engine, marketplace or assistant.
This pass read and diagnosed. The same structural understanding powers what follows — turning a diagnosed catalog into one where every colorway can be found by name.
1 · Apply the corrections the record already answers for itself
Name each colorway in its title and page title from its own option value. Give product type a real category and keep the bag family in its own field. Clear editor code from descriptions. Approved as patterns, not product by product.
2 · Settle the questions that gate the rest
Whether the checkout tools belong in the feed, which old addresses get a redirect, whether the apparel label is its own brand, which tags still drive a collection. Each is a judgment the data can show but cannot make.
3 · Hold the line at publish
Keep ongoing intelligence where the next drop is published, so a copied listing that still names the last colorway is caught before a shopper finds it.
This is how EKOM moves a catalog from insight to impact —
and keeps every colorway findable as the catalog grows.