A multi-lens data-quality read of the catalog — and where it leads.
This is a real EKOM catalog analysis, with the retailer's identity removed. The client is a specialty footwear retailer carrying running, hiking, comfort, and kids' footwear across established athletic and outdoor brands. EKOM ran a full-catalog quality read with no schema supplied and surfaced 62 flagged items — 57 of which are real, actionable catalog issues, the kind standard validation misses, because it checks whether a field is filled, not whether the value is right.
What follows leads with the findings a shopper hits first — customer-visible, live on the site today — then traces the pattern back to its structural root: a retailer-internal routing code that cross-wires gender and style across the catalog. Third-party brand and model names are kept as they appeared; only the retailer's own identity has been removed.
Where the 57 actionable findings concentrate, by theme. Many issues span more than one theme, so theme counts are approximate.
| Theme | ~Count | Who feels it |
|---|---|---|
| Gender & routing-code errors | 6 | Shoppers + ops |
| Gender-filter leaks | 9+ | Shoppers |
| Brand fragmentation & mistags | 2+ & catalog-wide | Shoppers + brand partners |
| Miscategorized product type | 5 | Shoppers |
| Broken commerce logic | 4+ | Ops |
| URL & handle errors | 5+ | Shoppers |
| Size & listing errors | 5 | Shoppers |
| Titles, copy & completeness | 5 | Shoppers |
Highest impact — the issues a shopper actually sees, live on the site today.
Two items from one athletic brand — plus at least one additional style elsewhere in the catalog — carry a rival brand's tag instead of their own. A shopper filtering by the correct brand never sees them. A shopper filtering by the rival brand sees shoes that brand didn't make — the kind of mismatch a brand partner notices in their own performance reporting.
A doubled-letter typo in two product handles produces a 404 for anyone who reaches them by search or share. A sandal's handle misspells one of its own spec terms; a hiking sandal's handle misspells its own colorway name. Small typos, but each one is a dead link on a live product.
Many items carry both an on_sale flag and a redundant "on sale" tag in the same record — the same signal stored in two places, with real risk of the two disagreeing if only one is ever updated.
One listing's spec bullets contain a raw, unformatted date string inserted mid-sentence — a data-pipeline artifact that made it onto the live page instead of staying in a back-office field. It reads as broken to anyone who lands on it, and it's the kind of defect a manual content review is unlikely to catch, because the rest of the bullet reads normally.
Most of what surfaces downstream — filter leaks, miscategorization — traces to one structural root.
The retailer's product feed carries an internal routing code that encodes gender and style together. Across several running-shoe models, the men's colorway carries the code for the women's version of the same shoe, and vice versa — six SKUs' worth, concentrated in a handful of styles. It's invisible to a shopper, but it's exactly the kind of mismatch that quietly breaks internal reporting and vendor-facing exports even when the storefront itself reads fine.
The same root shows up again, one layer up: four unisex sock items are filed as women's despite their own spec explicitly stating unisex sizing; a women's sandal and a women's slipper are both missing the gender prefix their sibling items carry; two items from one comfort brand use an ungendered product-type value instead of the gender-prefixed one used elsewhere in the catalog; and a unisex kids' item is classified under a boys-only category despite carrying both boys' and girls' tags. Each of these quietly narrows who can find the product in gender-filtered navigation.
Separately, two footwear brands are split by casing or naming across records — inconsistent capitalization on one, an inconsistent house-brand suffix on the other — so a brand-filtered collection silently drops part of its own inventory. Catalog-wide.
A closed-back clog style is filed under the wrong sandal category on two items; an elastic-closure sneaker from a comfort brand is miscategorized as an oxford despite every product signal indicating a slip-on; and a non-standard category tag appears on two more clog items instead of the catalog's standard form.
Five of the 62 flagged items are not catalog defects at all — they're a side effect of how EKOM's own pipeline prepared the data for review. Calling them out here, plainly, is part of keeping this analysis honest.
| Field | Severity | Why it's flagged — and why it isn't a retailer issue |
|---|---|---|
| tags | High | EKOM's per-product analysis view joins each product's gender, category, brand, style, and internal routing code into a single tags string for review. The pipeline correctly flags that string as "not atomic" — but the underlying catalog already stores these as distinct values; the joining happened in EKOM's own analysis layer, not the retailer's data. |
| product_type | High | Same pattern: EKOM's flattening combines gender segment and footwear style into one field for the product-level view used in this analysis. |
| option1_values | High | Size options across all of a product's variants were joined into one cell so the analysis could run at one-row-per-product. Flagged as non-atomic; a consequence of that flattening, not the source data. |
| option2_values | High | Same pattern, for color options across variants. |
| details | High | Spec bullets were joined into a single cell for the same reason. Flagged as combining several kinds of information — a byproduct of the analysis grain, not a defect in the live listing. |
Why a full read catches what a sampled check misses.
A sampled spot-check would have caught some of this and missed the rest — especially a routing-code pattern concentrated in a handful of styles. Reading the full catalog is what surfaced it cleanly, and it's what makes the next step precise: EKOM knows exactly which records need correction and which are already sound.
This pass read and diagnosed. The same structural understanding powers the work that follows — turning a diagnosed catalog into one that reads right everywhere it's seen.
This is EKOM's first read of this retailer's catalog. The pipeline profiled the catalog with no schema and no manual setup — identifying the vertical and every field's role before any analysis ran — then read the full live catalog rather than a sample, so a pattern concentrated in a handful of styles surfaced cleanly instead of averaging away. The findings here are that read, with the retailer's identity removed and third-party brand and model names preserved.