Resolution Layer Case Study · Read & Diagnose
A precision hand-tool manufacturer.
What a family-by-family read reveals — and where it leads.
This is a real EKOM catalog analysis, with the client's identity removed. The client is a precision hand-tool manufacturer — bits, drivers, sockets, keys, pliers, and torque tools. EKOM read the catalog the way it's actually built — split into tool families — and analyzed each in full. The headline isn't volume: on the whole, this is a well-run catalog. The point is what survives in a well-run catalog — a small set of high-consequence defects sitting in exactly the places standard validation and manual QA can't see: legal-compliance gaps, a pricing/identity collision, images that break on retail-partner pages, and unit errors that quietly make products unfindable.
These are machine-surfaced findings meant as a triage signal, not a verified defect list — a small share may be intentional or may resolve against the source system. Even read conservatively, the live critical items alone warrant action.
What's inside
- The findings that carry real risk — the live compliance, pricing, and channel defects most likely to cost money or a rejected submission.
- Why these were invisible — how the most consequential findings hide in the specialized lines a broad or sampled read can't see.
- How EKOM reads a catalog — why the catalog chose its own split, and why that surfaces family-specific defects.
- What it means, and what's next — from diagnosis to an enriched, channel-ready catalog.
Standard catalog validation checks whether a field is populated, not whether its value is correct. A torque screwdriver can pass every completeness check while stating an accuracy that contradicts its own description, or an inch size wearing a metric unit. EKOM's resolution layer reads structure and meaning together, family by family — which is why these surface here rather than passing straight through to a channel.
The weight is in severity, not count: 16 issues are critical and live right now. Many issues span more than one theme, so theme counts are approximate.
By tool family
| Tool family | Issues |
|---|
| Sockets | 15 |
| Screwdriver | 12 |
| Plier | 11 |
| L-Key | 11 |
| Torque Screwdriver | 8 |
| Bit | 7 |
By theme
| Theme | ~Count | Who feels it |
|---|
| Missing required fields (warranty, safety status, material, packaging) | ~15 | Retail channels |
| Wrong units & impossible measurements | ~13 | Shoppers + ops |
| Conflicting or wrong attribute values | ~13 | Shoppers |
| Set / single misclassification & component mismatches | ~12 | Shoppers + channels |
| Images that won't render (video URLs, TIFF) | ~5 | Retail channels |
| Compliance gaps (CA Prop 65) | ~3 | Retail channels |
| Identity collisions & self-references | ~3 | Ops |
The findings that carry real risk.
The issues most likely to cost money, trust, or a rejected submission — live in the catalog now.
Live compliance & commercial risk
A handful of defects with outsized consequences.
Two 8-piece impact socket sets are missing the CA Prop 65 warning required for metal tools sold in California — a legal-exposure and rejection risk. Two separate products share one identity (the same internal product ID and UPC) carrying conflicting prices and images, so downstream systems can't tell them apart. And a Precision Ball End Hex Screwdriver is stored at 6 lbs — a decimal slip; the true weight is 0.06 — which throws off freight and packaging math.
None of these are visible on a spot-check of the product page. They surface only when the catalog is read field-by-field, in family context.
Rejected before they reach the shelf
The fields retail partners validate against — left blank or unusable. A YouTube video URL sits in the image field on MicroBit, screwdriver, and hex-key items (with a blank hero image), and a Torx L-Key set uses TIFF files most channels reject at ingestion. Warranty Information is blank on ~30 screwdrivers and many socket and hex-key items; Safety Status is blank on socket and bit items. Each is a required display or validation field — present-but-blank passes a completeness check and still bounces the submission.
Invisible to the buyer
Unit and classification errors that quietly hide products. Inch sizes labeled metric — SAE hex keys at 0.050mm, retaining-ring pliers with mm on an inch Drive Size (a ~25× error) — land products in the wrong size facet. A 4-piece torque kit tagged Single never appears in bundle filters; a socket set tagged hex inch disappears from inch-socket search. The product is in the catalog, correctly priced — and unfindable by the buyer looking for it.
Why these were invisible.
The most consequential findings live in the specialized lines a broad or sampled read can't see.
The sparse-family effect
The Plier and Torque families alone held 19 issues a random sample surfaced none of.
The catalog is a block matrix — each family fills its own attributes (torque specs only on torque tools, jaw dimensions only on pliers) and leaves the rest blank. Those attributes appear on just 2–4% of products, so a random sample barely touches them. Reading each family on its own brought them into view: a torque handle stating ±10% accuracy against its own ±6% spec; conflicting torque-unit vocabulary; transposed plier dimensions; Material blank across a whole torque-driver line. These are the specialized, high-margin lines — precisely where attribute accuracy decides placement, and precisely where enrichment has the most upside.
A quiet proof of thoroughness: the first pass, a random sample, surfaced 14 issues and none from these families. Reading the catalog the way it's built surfaced 64.
This is the difference between checking a catalog and understanding one. A completeness check confirms a field is filled; it can't tell that a torque spec contradicts itself, or that an inch size is wearing a metric unit, or that the blank you're not looking at is the one a retail partner requires. Reading family by family is what makes both the defect and the fix legible.
How EKOM reads this catalog.
Why family-specific problems surface here rather than averaging away in a whole-catalog scan.
1
Profile the catalog
EKOM profiled the catalog with no schema supplied — inferring the vertical and every field's role, fill rate, and quality signals, and detecting that attributes cluster by tool family.
No manual setup required.
→
2
Choose the split & pass
It measured which field unlocked the most sparse-but-family-dense attributes. Tool Type unlocked 17; Country of Manufacture, just one. So it split by Tool Type and ran a focused pass per family.
The catalog chose its own split.
→
3
De-duplicate & group
Findings from every family are unioned onto the single catalog, duplicates removed, then grouped by theme and by who feels it.
Signal, not noise.
Reading each family in full context is how a torque spec that contradicts itself gets caught instead of buried. It's also what makes the next step precise: because EKOM knows which attributes each family actually needs, the enrichment that follows targets the right fields for the right products — not a generic fill.
more than
25%
More than one in four organizations facing data-quality challenges report losing over $5 million annually as a result — and for a manufacturer, that figure doesn't capture the compounding cost in the channel, where every retail partner's search filter, size facet, and validation gate relies on clean structured data. Catalog defects don't announce themselves; they surface as rejected submissions, products missing from a partner's filters, and specs a buyer can't trust.
Forrester Data Culture and Literacy Survey, 2023 (RES181258)
The consequential defects concentrate in the specialized lines — torque, pliers — where a handful of wrong or missing attributes decide whether a product is compliant, findable, and accepted by a channel. That's not a sign of a neglected catalog; it's the nature of a good one. The defects that survive careful cataloging are the ones no completeness check is built to catch, and they hide in the highest-value families.
None of these are content problems. A torque handle with a blank Material never appears in a material-filtered search no matter how well it's described. A missing CA Prop 65 warning is a compliance and rejection risk no product copy can fix. An inch size labeled in millimeters puts the product in the wrong facet, invisible to the buyer looking for it. Standard validation checks for presence; EKOM's resolution reads for correctness — reading each family in full.
This pass read and diagnosed. The next step is where the catalog gets better, not just cleaner.
From diagnosis to channel-ready.
The read in this report is the first phase. The same family-by-family understanding that surfaced these findings is what powers the work that follows — turning a diagnosed catalog into one that wins placement across every channel.
1 · Complete the family-specific attributes
Normalize and fill the sparse torque, plier, and socket fields that channels validate against — units, materials, torque specs, safety and compliance flags — so every product clears every partner's ingestion gate the first time.
2 · Enrich for discovery
Richer, standardized attributes and content tuned for on-site search, marketplace facets, and AI answer engines — so products don't just pass validation, they show up and win the click across the channels that carry them.
3 · Stay channel-ready as the line grows
Ongoing catalog resolution so new products and updates stay clean, complete, and syndication-ready — the understanding compounds instead of eroding with every new SKU.
This is how EKOM moves a catalog from insight to impact —
and from a clean catalog to a channel-winning one.
About this analysis
This is EKOM's second pass on the same catalog. The input did not change — the method did. The first pass read a random sample and surfaced 14 findings. This pass read the catalog by tool family, and required no schema and no manual setup — the pipeline profiled the catalog on its own, chose the split axis from the data, and read each family in full context, surfacing 64 findings. Same catalog, read the way it's built, with nothing configured by hand.