The Resolution Layer · Case Study in BriefEKOM
Electrical tools & test equipment.
A cold read of a sample of the tools and test-equipment product pages of an electrical distributor — where a buyer chooses an instrument on its type, its voltage and its safety category before anything else. One of a set of anonymized EKOM analyses — what surfaced, and where it leads.
Lead finding · instruments with a brand and nothing else
20 product pages in the sample whose specification table has one row: Brand.
Eighteen are instruments from one test-equipment maker — a cable certifier, optical loss test sets, an insulation tester, clamp meters — and two are carts from one cable maker. An electrician filters on type, voltage and safety category; these pages carry none of the three, so they drop out of the type, voltage and safety-category filters and the comparisons built on them, while products with more specification fields stay in. Other pages from the same instrument maker in the sample carry the fields, so the gap follows the record, not the manufacturer. On a test instrument the missing fields are the whole decision.
Also surfaced
One product, three brands
A compartment case whose browser-tab title names one maker, whose heading names another, and whose specification table names a third. A purchasing agent who sees two makers stops trusting the rest of the page.
Misspelled values
Product types and names with a letter wrong — each one a search or filter that misses the product, and for most of them the correct spelling is already on a sibling item.
One fact entered two ways
In the sample, multi-color values run together on 23 pages and slash-separated on 7, so one color combination splits across two filter values; all 4 drive sizes in the sample carry no unit.
A sitemap that stops short of the catalog
Some departments get one product address each, while a large block of addresses sits under a discontinued-products branch.
How EKOM read it. The engine read a sample of 117 public product pages cold across 25 subcategories — no file, no credentials, no integration — plus the product sitemap, then had an analyst re-check the findings against the saved pages. It profiled what each field is for from how the catalog itself uses it, then held every page against its neighbors: a brand against its title, a size against its unit, a color against its siblings. The extents of 20, 23, 14 and 7 are an analyst's recount within the sample, not the automated pass's own count.
From diagnosis to channel-ready. The 11 findings and four site-wide checks read mostly as five upstream mechanisms — one page template behind three site-wide symptoms, a manufacturer data load that reached some records and not others, a sitemap generator that stops early, free-text entry with no check against the catalog's own vocabulary, and one field made to carry several meanings. Fixing the findings fixes the findings; fixing the mechanisms fixes every page that has already passed through them and every one that will. Then the same intelligence holds the line at intake, so the next misspelled type is caught the day the product is listed.