A sales report becomes more useful when it speaks the store's own vocabulary. Fabric, designer and supplier can matter more to a buying decision than Shopify's standard product categories.

What actually happened

On 17 November, Shopify expanded Sidekick analytics to use metafields and metaobjects alongside standard store performance data. A merchant can ask a question such as total sales by fabric in the previous quarter and receive a breakdown based on the product's custom attributes.

The feature is not automatic for every custom property. Shopify says supported metafields need the 'Filter or group data in Analytics' option enabled, and some metafield types and reports are not yet compatible. Sidekick uses the information already maintained in the store rather than filling in missing attributes.

Why a small team cares

Imagine a clothing store with several hundred products, each tagged with material in a metafield. A sales-by-fabric breakdown could show whether linen or cotton performs better in a particular season and help the owner plan their next order.

Previously, this could mean exporting sales and product records, joining them in a spreadsheet and rebuilding the result each month. Sidekick creates a faster path from a concrete question to an initial view, particularly for teams without a data analyst.

The value is not asking more questions. It is making one better buying or merchandising decision with less reporting effort. The same principle works for supplier or designer metadata when those attributes are supported and maintained consistently.

Hype vs useful

An AI response does not repair unreliable store data. Inconsistent values such as 'cotton', 'Cotton' and '100% cotton', or incomplete attributes, can distort the result or leave part of the catalog unclassified. Supported field types also impose real limits.

For a store with clean existing attributes, this is ready to adopt in a controlled reporting workflow. Check the aggregation against Shopify's conventional reporting before changing purchasing plans. If the underlying fields are inconsistent, cleanup is the first task, not prompt engineering.