The Resolution Layer  ·  Case Study in BriefEKOM

Sports & fitness equipment.

A full-catalog read of a sports and fitness equipment retailer — where the storefront shows a shopper one thing and the product feed tells every carrier, marketplace and assistant another. One of a set of anonymized EKOM analyses — what surfaced, and where it leads.

Vertical
Sports and fitness equipment retail — national brands alongside an own-brand range
Method
Full-catalog read · auto-profiled, nothing supplied
Issues surfaced
104 actionable (115 flagged) · 31 critical
This pass
Read & diagnose
Lead finding  ·  the page and the feed disagree
The page says "Price on Request." The feed gives a five-figure price.
159 variants carry the same round five-figure number in the price field, every one of them marked unavailable, most of them commercial equipment sold by quotation. On the storefront the retailer handles them well — the page reads "Price on Request." The feed publishes the number, and a shopping engine, a price-comparison site or an assistant reads it as real. One brand accounts for 139 of them, and some records are plainly not five-figure goods at all. The fix is one decision, applied once: how a quote-only product should appear in a feed.

Also surfaced

Shippable goods that weigh nothing
606 variants need shipping and carry a weight of zero — 252 of them priced at a hundred or more, including commercial machines and a four-figure upright bike. Anything that rates delivery by weight is rating them on nothing. In several cases the weight is already stated in the product's own details.
Content that belongs to the product next door
Two attachments wearing each other's names, a tennis net carrying volleyball copy, a goal whose size option names its larger sibling, photos of the thinner mat and the smaller goal. The wrong value is usually a real value belonging to another product.
Records that contradict themselves
A 24kg dumbbell set with its weight typed in grams, an exercise ball sized in metres, a latex warning cut off mid-word, a rack titled as one product and specified as another. Every field is filled; the record still gives two answers.
Identity and addresses
Twelve saleable lines with no SKU, most of them bundles. A competing brand as a vendor. Ten findings where a permanent URL names a different spec from the title — including an upright bike at a recumbent bike's address, with "-copy" on the end.

How EKOM read it. The engine took the complete public storefront feed — no file, no credentials, no schema — profiled the catalog's own field semantics, then held every variant against its siblings, its title and its own details. An analyst then recounted the headline extents across the whole catalog from the archived feed, which matched the analyzed table on every row. That is what separates a feed defect from an extraction artifact: here there is no gap between the two to hide in.

From diagnosis to channel-ready. The 104 findings collapse into seven likely upstream mechanisms — "price on request" written as a number, products created by duplicating a neighbour, weight held in the copy instead of the field, bundles missing what single products carry, addresses never updated, free text where a controlled value belongs, and tags doing three jobs at once. Fixing the findings fixes the findings; fixing the mechanisms fixes every product that has passed through them and every one that will. Then the same intelligence holds the line at intake, so the next duplicated or weightless product never reaches the feed.

The Resolution Layer
EKOM
[email protected]  ·  ekom.ai
EKOM
Case study — client anonymized