Variant exports: define models and row grain
Keep shared product models, individual variants and source identifiers distinct. Explore a Portwest reference and a fictional variant export.
iWeb describes rebuilding Portwest’s product model during an Akeneo migration. Its account says the model was organized around variant structure, attribute groups, completeness checks and channel requirements, with product identifiers supporting downstream assets and catalogue output. It reports checking migrated data and output dependencies before the live cutover. This is the implementation partner’s account, including customer commentary. It does not publish a complete variant export or the rules for every identifier and variation axis.
State what one row represents
For a marketplace-data export, PandaBuyNow proposes defining the row’s entity type before comparing its values. A shared product model and an individual variant need distinguishable records or clearly named columns. Retain the source listing identifier, available variant identifier and source URL alongside any local row key.
Document which attributes describe the shared model and which distinguish variants. Preserve the observed option labels and their source context. If the source provides no variant identifier, mark that absence in the data-quality notes; a local row key should remain identifiable as a delivery reference.
Test the structure with fictional variants
Consider invented model JACKET-Q with three observed variants: navy/S, navy/M and black/M. The shared listing identifier can appear on all three rows while their row keys and option combinations differ. Deduplicating on the listing identifier alone would remove valid variant records.
The observed combinations do not establish a black/S variant. Keep that combination absent unless the source supplies it. These labels illustrate the proposed export and describe no Portwest or PandaBuyNow product.
Check the delivered structure against the agreed sample
Ask the recipient to check a shared identifier, a missing variation axis and a variant whose parent reference cannot be resolved. Record the agreed uniqueness rule and the treatment of unresolved links in the field definitions. A JSON structure may separate models and variants; a CSV can carry explicit entity and parent-reference columns. The chosen format should preserve the same relationships. Include the row count by entity type in the delivery notes so the recipient can detect an unexpected loss during import.
For a PandaBuyNow discussion, provide representative marketplace links, the required row structure and target fields. Agree the JSON or CSV sample, source references and data-quality notes before expanding the data delivery. The recipient can use that file in its own catalogue system.