Resolution Layer Case Study · Read & Diagnose
A national furniture retailer.
An end-to-end read of a marketplace-bound catalog — and where it leads.
This is a real EKOM catalog analysis, with the retailer's identity removed. The client is a national furniture retailer whose catalog feeds a third-party marketplace, so every record has to clear platform validation before it can sell. EKOM ran the catalog end to end — auto-profiled, with no hand-built schema and no manual setup — and surfaced 72 specific, actionable defects: impossible material math, country-of-origin contradictions, blank identity keys, Excel-corrupted dates, non-compliant Prop 65 placeholder text, understated kit counts, and pervasive marketplace-validation failures.
These are machine-surfaced findings meant as a triage signal, not a verified defect list — a small share may be intentional. Even read conservatively, the ones that map to marketplace rejection, legal exposure, and fulfillment error warrant action.
What's inside
- The findings that carry real risk — impossible material math, origin contradictions, and compliance placeholders that map straight to money.
- Marketplace-validation failures — the required fields left blank or malformed that get a listing rejected before it can sell.
- How EKOM reads a dense catalog — the granularity ladder, and why small focused passes out-find a single sweep.
- What it means, and what's next — from diagnosis to a clean, marketplace-ready catalog.
Standard catalog validation checks whether a field is populated, not whether its value is correct. A dining set can pass every completeness check while its material percentages sum to 300%, its origin contradicts its shipping port, or its Prop 65 label still holds a template placeholder. EKOM's resolution layer reads structure and meaning together — which is why these surface here rather than at a marketplace rejection or a customs desk.
A dense, marketplace-bound catalog: each record carries dozens of attributes that must pass platform validation, so the defects are per-item and sharp.
The count isn't the point. These are defects that pass a completeness check — the field is filled — yet are wrong in ways that only surface when structure and meaning are read together. They map to three costs: marketplace rejection, legal & customs exposure, and fulfillment error.
The standout findings
- Impossible material math. A Faybrid table lists Aluminum three times, summing to 300%; two Armen Living items have material-content percentages exceeding 100%.
- Origin vs. factory contradictions. Sagebrook items list Canada as origin while shipping from Chinese/Indian ports; Martin Furniture / JLA items are tagged Domestic despite Vietnam/Indonesia factories — customs and country-of-origin marking risk.
- Blank identity key. The designated product_sku is blank on nearly all items, with the SKU value sitting unused in another field — records unlinkable to purchase orders and warehouse systems.
- Prop 65 placeholder text. Warning labels contain unfilled template brackets instead of chemical names — not legally compliant.
- Excel epoch corruption. First-inventory dates of 12/31/1899 across many items.
- Kit shortfalls. Multi-piece sets declare fewer components than the product name states — one standalone chair is named as a 9-piece set.
Where it fails marketplace validation.
The single most consequential pattern for a marketplace-bound catalog: required fields present-but-blank or malformed, so the listing is rejected before it can sell.
Pervasive validation failures
Required fields blank — the listing can't go live.
Across the catalog, the platform flagged records INVALID for missing LegMaterial, ProductName, Height, ProductSize, and carton dimensions. Eleven Sagebrook items carry a web/SEO title but a blank required product name; LegMaterial is blank on nine-plus items across vendors — an invalid record and a missing filter facet at the same time. Each blocked listing is a product that exists, is priced, and cannot be bought.
A completeness check that only counts populated fields waves many of these through; the marketplace's own validator does not.
Fulfillment & freight blockers
Missing Height and carton dimensions break freight classification and quoting on pub tables, dining chairs, and bed frames; a shippable item lists Volume = 0; two bed frames are typed KIT with no component definitions — unfulfillable as listed. Downstream, blank product_sku means those records can't be linked to a purchase order or warehouse system at all.
Pricing & routing integrity
Several active dropship products show $0 local and dropship cost — a margin-calculation error; one pub table has a blank dropship cost; and ProductMiraklProvider is blank on items that then can't be assigned for routing or commission. A special-order lead time of 1000 days (~2.75 years) sits on multiple items as an unremoved placeholder.
How EKOM reads this catalog.
Small enough to read whole, but dense with per-item defects — so a single sweep averages the sharp ones away.
1
Profile the catalog
EKOM auto-profiled the catalog with no schema supplied — identifying the vertical, every field's role, fill rate, and quality signals — and chose its analysis strategy from the data.
No manual setup required.
→
2
Run the granularity ladder
One broad whole-catalog pass for catalog-wide patterns, plus three focused passes for sharp single-item defects. The choice of ladder over a category split is deterministic for a catalog this size.
Broad + focused, together.
→
3
Group by impact
Findings are consolidated and grouped by risk — marketplace rejection, legal/customs exposure, and fulfillment error — so the fix is prioritized by cost, not by field.
Signal, not noise.
Why the ladder matters
Each focused pass out-found the broad one.
The three focused passes surfaced 18, 23, and 16 findings — each more than the broad pass's 15. A small, attentive pass catches per-item defects a large sweep misses: impossible material percentages, $0 costs, swapped or contradictory attributes, and Excel epoch dates only surface when each record is read closely, in a pass sized to see it.
What this means — and what's next.
more than25%
More than one in four organizations facing data-quality challenges report losing over $5 million annually as a result — and for a marketplace-bound furniture catalog, that cost is concrete: a rejected listing doesn't sell, a Prop 65 placeholder is legal exposure, a mislabeled origin is a customs risk, and an understated kit count is a return waiting to happen.
Forrester Data Culture and Literacy Survey, 2023 (RES181258)
None of these are content problems. A dining set whose material percentages sum to 300% passes every "not blank" check and is still factually impossible. A blank product_sku populates nothing a shopper sees and quietly severs the record from every operational system behind it. A Prop 65 label with an empty placeholder reads as complete and is not compliant. These accumulate because standard validation checks for presence, not correctness — and a marketplace's validator, unlike a completeness check, rejects them.
This pass read and diagnosed. The same structural understanding powers the work that follows — turning a diagnosed catalog into one that clears validation, ships correctly, and stays compliant as the assortment grows.
1 · Apply the confirmed corrections
Resolve the impossible material math, origin/factory contradictions, kit shortfalls, and Prop 65 placeholders, and populate the identity key — the defects EKOM can correct from data already in the catalog.
2 · Clear the validation gate
Fill the required fields the marketplace rejects on — LegMaterial, ProductName, Height, ProductSize, carton dimensions — so every listing passes validation the first time and appears in the right filters.
3 · Hold the line at intake
Ongoing catalog resolution so new vendor loads come in clean, compliant, and marketplace-ready — the understanding compounds instead of re-accumulating the same gaps with every new SKU.
This is how EKOM moves a catalog from insight to impact —
and from a validation backlog to a catalog that sells.
About this analysis
This is EKOM's second pass on the same furniture catalog. The input did not change — the engine did. The first pass was graded against a supplied marketplace specification; this pass auto-profiled the catalog on its own — vertical, field roles, fill rates, quality signals — and chose its own analysis strategy, with no schema supplied. Where the first pass mapped what was missing, this one surfaced what is present but wrong: impossible material math, origin contradictions, placeholder compliance text, and Excel epoch dates that pass every coverage check. It ran a four-pass granularity ladder (one broad, three focused), and each focused pass out-found the broad one. Same catalog, read deeper, with nothing built by hand.