This worked case is part of the catalog correction workflows guide, which covers what a platform consolidation can leave behind. Here the job is different: follow one hard case from start to finish, mark who decided each step and where the evidence lives, then turn that into signs a marketplace operations consultant can recognize at a client.
The case: a dinner set that was both stoneware and ceramic
Illustrative. A fictional 140-person housewares brand sells through its own store and two marketplaces, with product content managed in a PIM. Nothing about the product was wrong. A 16-piece dinner set is stoneware, serves four and was correctly described in every field. The record was inconsistent: the material attribute said "Stoneware", while the title, generated from a family template, said "Ceramic Dinnerware Set". The item count said 16 while a bullet said "service for 4". Each statement was true. Together they tripped the marketplace's consistency checks.
That is the exception this page is about: a compliant product rejected for inconsistent attributes. It matters because the fix is not obvious. A specialist could change the product, the title, the attribute or the template, and each choice has consequences for other products.
Step by step: who decided what, and which record shows it
| Step | Date | What happened | Who decided | Record that captured it |
|---|---|---|---|---|
| 1 | February 3 | Version 2 created: material "Stoneware", template title "Ceramic Dinnerware Set" | Catalog specialist | PIM version with author, timestamp, old and new values |
| 2 | February 4 | Marketplace returns an error: material does not match title; listing not updated | Marketplace checks | Issue code and attribute named, saved in the brand's operations tracker |
| 3 | February 5 | Template, not product, will change: stoneware family titles will use the material attribute | Catalog operations head | Tracker assignment and decision note |
| 4 | February 6 | Version 3: title template and item count wording updated | Catalog specialist | PIM version log for each changed field |
| 5 | February 7 | Resubmitted; error clears; listing goes live | Marketplace checks | Listing status and resubmission date |
| 6 | February 20 | Version 4: template change rolled out to 38 other stoneware products; sign-off | Catalog operations head | Template change log and bulk version entries |
| 7 | Early March | Check shows no new material mismatch errors on the family | Catalog operations head | Tracker filtered by family and issue code |
Seven steps, three people or systems making decisions, and three places where the evidence lives: the PIM, the marketplace, and the brand's own tracker.
Why the decision at step 3 is the valuable part
Step 3 is where a person chose between options. Changing one product's title would have cleared this listing and left 38 others exposed. Changing the material attribute would have made the data wrong. Changing the template fixed the cause. The tracker note recording that choice, and the operations head's name on it, is the piece of evidence nobody can reconstruct later.
The PIM side tends to be well kept. Akeneo, for example, records for each product version the author or system, the timestamp and the old and new values, with versions created for attribute, status, family, association and category changes. Steps 1, 4 and 6 are therefore usually recoverable, as long as version history survives.
The marketplace side is weaker. Amazon's Selling Partner API returns a listing's existing issues when issues are requested, and ERROR severity issues block the update while WARNING issues do not. Issue changes are reported through notifications. An interface that reports the issues present now is not an archive of every rejection the brand ever received. Step 2 exists only because the brand saved the issue in its tracker on the day.
The outcome side is easy to lose too. If exported feed files overwrite each other, as Akeneo notes scheduled exports can unless a naming pattern is set, there is no snapshot of the listing before and after. Step 5 then depends on a resubmission date somebody wrote down.
What makes this sequence useful evidence
Three properties make a case like this worth noting, and they are the same ones to look for in any client's history.
- An initial problem with a recorded reason. Not just "listing failed", but the issue and the attribute named, captured at the time.
- An authorized action. A named person decided the fix, and the decision is visible: the template, not the product, was changed.
- A checked result. The resubmission cleared, the fix was rolled out, and a later check showed no recurrence.
Records with all three show cause, decision and verification in order. For AI developers training systems to handle catalog operations, that chain shows something clean product records cannot: how an experienced team reasons about a conflict between rules. The parallel guide on human review records and correction workflows describes the same pattern more generally, and logistics exceptions and their documented resolutions shows it in freight.
How to recognize similar cases at a client
A consultant can usually tell from a short conversation whether a client has cases like this one. Listen for these signs:
- Rejections are tracked somewhere. A tracker, ticket queue or shared sheet lists issue codes against SKUs. Without this, step 2 is missing.
- Templates or rules drive titles. If titles are generated from attributes, inconsistency cases are common and their fixes ripple across families.
- Version history is switched on and kept. Ask how far back the PIM history goes, and whether a previous migration reset it.
- Someone signs off on bulk changes. A named approver for template or family changes means step 6 is attributable.
- The team checks whether fixes held. A habit of reviewing issue counts by family after a change gives the checked result.
- There is a real end date. A PIM being retired, a seller account changing hands or an agency contract ending puts the history at risk.
Note only the metadata: which systems, how many years, roughly how many rejections, who approves. Do not ask for product files, tracker exports or seller account access. To map how the systems join, use the records across systems map. To find the person who can approve a discussion, use the who can license tree.
Qualification and what stays open
The company needs 50+ full-time employees at peak (contractors excluded), a documented operating history, rights to license the records, and an authorized contact. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English.
SourceX licenses business data directly from companies and pays them for it. We are their counterparty from agreement through delivery and payment. How buyers generally approach data is outlined in how AI labs buy data; nothing here implies a particular buyer. More sector context is on the ecommerce industry page.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
If a client can follow a rejected listing from issue code to accepted fix, register as a partner, check the company on who qualifies, and read the program terms before you introduce the catalog operations head.