E-commerce &
retail marketing
← The field guide

Product data research

Reproduce studies of product pages and published category mappings. Read the methods, inspect the results, and download the code to repeat each analysis.

George Kelly · Observed

Product dimensions in structured data across four retailers ↗

On November 14, 2026, all 12 sampled U.S. furniture listings showed overall product measurements. Six included product dimensions in their Schema.org markup. Five matched the visible specifications exactly across three named axes. The sixth used different labels for one axis. Six listings showed package dimensions to shoppers, and the captured Schema.org markup contained package dimensions on zero of the 12.

Study files

George Kelly · Observed

Shopify to Google product category mapping across three releases ↗

Shopify’s November 2026 mapping sends 63,924 Shopify category IDs to 5,455 distinct Google category IDs. In that release, 62,848 Shopify IDs share a Google destination with at least one other Shopify ID. The exported rules show where a feed can lose category detail. Merchant Center’s assigned category still needs to be checked on the processed offer.

Study files

George Kelly · Observed

Comparing product dimensions across country pages: a 30-listing audit ↗

Of 30 sampled US listings, eight had retrieved same-article US and GB records with three comparable named product axes. None matched exactly after conversion; all eight fell within the preselected 2 mm display-comparison threshold. Twenty comparisons remained unresolved because of retrieval gaps, and two because the pages reported diameter and height rather than three named axes. These are limits of this observation method, not an estimate of product-data error rates.

Study files

George Kelly · Observed

Product variant URLs: what eight public pages actually selected ↗

In this eight-variant public sample, every exact URL selected the intended options and exposed a matching structured Offer. Reading the first Offer would still attach the wrong SKU in 5 of 8 cases and the wrong price in 2. Resolve the selected variant before choosing its offer; a common canonical URL or equal sibling price cannot do that work.

Study files