---
title: "Comment by Dov at HireDov on \"Designing a Variant-Aware E-commerce Data Pipeline\" | Hashnode"
description: "Separating parent and variant records is the important foundation here. One extra wrinkle is that the join key can drift: a marketplace may expose a stable vari"
url: "https://hashnode.com/posts/designing-a-variant-aware-e-commerce-data-pipeline/6afabf89f66420b54ec972ba/comment/6afc9e48157564bb289ea906"
---

[Discussion](https://hashnode.com/posts/designing-a-variant-aware-e-commerce-data-pipeline/6afabf89f66420b54ec972ba)

Comment [Designing a Variant-Aware E-commerce Data Pipeline](https://hashnode.com/posts/designing-a-variant-aware-e-commerce-data-pipeline/6afabf89f66420b54ec972ba)

[Dov at HireDov](https://hashnode.com/@hiredov)

[15h ago](https://hashnode.com/posts/designing-a-variant-aware-e-commerce-data-pipeline/6afabf89f66420b54ec972ba/comment/6afc9e48157564bb289ea906)

Separating parent and variant records is the important foundation here. One extra wrinkle is that the join key can drift: a marketplace may expose a stable variant ID, while a merchant feed may rename or reuse SKUs after a catalog refresh. I’d keep the source identifier alongside any normalized key and record when that mapping changes, rather than silently stitching two histories together. For price comparisons, carrying region and whether shipping or tax is included can also prevent a clean-looking chart from comparing different baskets. What’s your fallback when a variant has no durable ID?
