The Resolution Layer · Case Study in BriefEKOM
A supplements and wellness brand.
A full-catalog, no-schema read across protein powders, capsules and topicals — sold alongside a large partner-linked range in the same feed. One of a set of anonymized EKOM analyses.
Lead finding · contradiction
Four products in a plant-based line describe themselves as containing a dairy protein.
The product name says plant-based. The copy underneath names an animal-derived protein source. A shopper choosing this product to avoid dairy is reading one of two statements and has no way to know which is right. The provenance is visible in the data: the brand's separate dairy-protein line publishes that same sentence, where it is accurate — the plant-based line was built from it and the copy came along.
Also surfaced
Four partner listings pointing at a fifth product's page
Five listings from one partner carry an identical destination URL, down to the variant id. On one it is correct; on the other four a shopper who clicked a specific item lands on something else — and attribution follows the link, not the listing, so the commission is credited against the wrong product. A separate listing carries an image file path where its destination should be.
Storefront URLs still naming the product they were copied from
At least seven products were duplicated and repurposed without the URL handle being changed, so the address a shopper sees and shares describes a different item. Five live titles still carry the draft artifact.
Two businesses sharing one feed, and the conventions crossing over
A tag reserved for stocked first-party inventory sits on at least nine partner rows priced at zero, and five of the brand's own products are classified under the partner vendor value. Neither is visible to a shopper. Both are visible to every report and outbound feed that trusts those fields.
Live products with nothing for a channel to read
Four recently launched products have no body copy at all — not thin copy, none — and four more carry prose but no structured highlights, the shape most channels ingest. For a brand whose case is the ingredient rationale, those pages ask a shopper to buy on the name alone.
How EKOM read it. With no schema and no manual setup, the engine profiled the catalog and read it in full rather than sampling. What it learned first was structural: this storefront runs two businesses in one product feed — the brand's own line and a large partner program — and they follow different rules. Most of what surfaced sits on one side of that line, which matters less for what it costs today than for who should fix it.
Eleven findings, five mechanisms. Each finding is downstream of a process wrong in one specific, repeatable way, and each leaves a signature in the data that points to it — copy inherited from a sibling product, duplication that leaves the identity fields behind, a destination field with nothing validating it, one business's conventions applied to the other, and two values on one record with nothing deciding between them. Correcting eleven findings corrects eleven. Correcting five mechanisms addresses every product that has passed through them and the ones that will. Five prepared corrections are derivable from the catalog itself; six questions need an answer that exists inside the business, and this read does not guess at them.